Science of the Total Environment 645 (2018) 630–645
Contents lists available at ScienceDirect
Science of the Total Environment
j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenvRainfall isotope variations over the Australian continent – Implications
for hydrology and isoscape applicationsSuzanne E. Hollins, Catherine E. Hughes ⁎, Jagoda Crawford, Dioni I. Cendón, Karina T. Meredith
Australian Nuclear Science and Technology Organisation, Locked Bag 2001, Kirrawee DC, NSW 2232, AustraliaH I G H L I G H T S G R A P H I C A L A B S T R A C T• Stable isotopes in precipitation decode
the hydrological cycle.
• Long term stable isotopes in Australian
precipitation at 15 sites presented.
• Relationships with meteorology/geog-
raphy used to develop continental
isoscape.
• Rainfall isotope and groundwater con-
nections established for NE Australia.E-mail addresses: Suzanne.Hollins@ansto.gov.au (S.E.
(D.I. Cendón), Karina.Meredith@ansto.gov.au (K.T. Mered
https://doi.org/10.1016/j.scitotenv.2018.07.082
0048-9697/Crown Copyright © 2018 Published by Elseviea b s t r a c ta r t i c l e i n f oArticle history:
Received 10 April 2018
Received in revised form 12 June 2018
Accepted 6 July 2018
Available online 18 July 2018
Editor: José Virgílio CruzThis paper presents a continental scale interpretation of δ2H and δ18O in Australian precipitation, incorporating
historical GNIP data at seven sites (1962–2002) and 8–12 years of new monthly data from 15 sites from 2003
to 2014. The more than doubling of stations and the significant time series duration allow for an improved anal-
ysis of Australian precipitation isotopes. Local meteoricwater lines were developed for each site, and for the Aus-
tralian continent. When the annual precipitation weighted values were used, the Australian meteoric water line
was δ2H= 8.3 δ18O+ 14.1‰.
Precipitation amount was found to be a stronger driver of precipitation isotopes than temperature at most sites,
particularly those affected by tropical cyclones and themonsoon. Latitude, elevation and distance from the coast
were found to be stronger drivers of spatial variability than temperature or rainfall amount.
Annual isoscapes of δ2H, δ18O and deuterium excesswere developed, providing an improved tool to estimate pre-
cipitation isotope inputs to hydrological systems. Because of the complex climate, weather and oceanicmoisture
sources affecting Australia, regional groupings were used instead of the climate zone approach and additional
data was included to improve the coverage in data poor regions. Regression equations for the isoscape were de-
rived using latitude, altitude and distance from the coast as predictor variables.
We demonstrate how this isoscape can be used as a tool for interpreting groundwater recharge processes using
examples from across Queensland and New SouthWales, including the Murray Darling Basin. Groundwater iso-
topes at sites where direct local recharge occurs are similar to rainfall, but for inland sites, which are often arid or
semi-arid, a disconnect between shallow groundwater and local rainfall is observed; the departure in deuterium
excess for these sites increases with aridity and distance from the headwaters where flooding originates.
Crown Copyright © 2018 Published by Elsevier B.V. All rights reserved.Keywords:
Deuterium
Oxygen-18
Precipitation
Local meteoric water line
Groundwater
GNIP⁎ Corresponding author.
Hollins), Cath.Hughes@ansto.gov.au (C.E. Hughes), Jagoda.Crawford@ansto.gov.au (J. Crawford), Dioni.Cendon@ansto.gov.au
ith).
r B.V. All rights reserved.
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 6311. Introduction
The stable isotopes of hydrogen (2H, or D) and oxygen (18O) are
commonly used as tracers of the hydrological cycle. For example, they
can be used in studies of surface water–groundwater interactions
(Kalbus et al., 2006), in the estimation of evaporative loss (Skrzypek et
al., 2015) and estimating groundwater recharge (Adomako et al.,
2010; Cartwright et al., 2017). In addition, the relationship between sta-
ble isotopes of precipitation with temperature (and rainfall) has been
used to interpret paleoclimate records (Rozanski et al., 1993;
McDermott, 2004; Vachon et al., 2010; Treble et al., 2013). In a large
number of cases, these relationships have been studied using the Global
Network of Isotopes in Precipitation (GNIP) data (e.g. Dansgaard, 1964;
Araguás-Araguás et al., 2000). More recently, efforts have beenmade to
predict the spatial variation in isotopic composition (isoscape) of pre-
cipitation for regions where little GNIP data exists: the Online Isotopes
in Precipitation Calculator (OIPC, Bowen and Wilkinson, 2002; Bowen
and Revenaugh, 2003; Bowen, 2018) and the Regionalized Cluster-
based Water Isotope Prediction (RCWIP, Terzer et al., 2013). As stated
in Terzer et al. (2013), this was due in part to the increasing demand
for spatio-temporal predictions of δ2H and δ18O values in precipitation
for use in ecological, wildlife forensics (Bowen et al., 2005) and food
source traceability studies after it was found that there was correlation
between the δ2H and δ18O values of some plant, animal and human tis-
sues and isotopic patterns of precipitation.
The isotopic composition of precipitation can be affected by itsmois-
ture source, followed by the condensation temperature and the precip-
itation history in the cloud (Dansgaard, 1964). Further, as the raindrops
fall below the cloud, it can be affected by moisture exchange with the
surrounding vapour and sub-cloud evaporation, which increases 18O
and decreases deuterium excess (defined as d = δ2H– 8 × δ18O;
Dansgaard, 1964; Froehlich et al., 2002). Thus the regional climate can
have a considerable impact on the isotopic composition of precipitation
at a particular site. As a result of these processes, depletion in 2H and 18O
occurs with increasing distance from the coast (continental effect) and
increasing elevation (altitude effect; e.g. Araguás-Araguás et al., 2000).
A latitude effect also exists (Feng et al., 2009), with a local minimum
near the equator and two maxima on either side, which coincide with
the subtropical highs. These processes can have an impact on the isoto-
pic composition of Australian regional precipitation, as a number of cli-
mate zones occur over the Australian continent, with tropical conditions
in the north, sub-tropical further south and then temperate conditions
in the south of the continent. Moving inland from the coast, we see
the vegetation distribution changes from woodlands or rainforest, to
grassland regions followedby deserts, reflecting the reduction in rainfall
with distance from the coast. Liu et al. (2010), in their study of Austra-
lian GNIP data from 1962 to 2002, noted that rainfall at the single inland
site in their study was isotopically more depleted than the coastal sites
and interpreted this as a continental effect.
The isotopic values of precipitation are also affected by synoptic
weather patterns (Barras and Simmonds, 2008; Scholl et al., 2009;
Baldini et al., 2010) with some of the lowest 2H and 18O values re-
corded in rainfall from tropical cyclones (Gedzelman et al., 2003).
Tropical cyclones are experienced in the northern half of the Austra-
lian continent. Continuous measurements of δ18O and δ2H in precip-
itation and water vapour were reported by Munksgaard et al. (2015)
during the passage of the tropical cyclone Ita in north-eastern Aus-
tralia. The Australian monsoon also dominates rainfall in northern
Australia. Zwart et al. (2016) undertook a study in northern Australia
during twomonsoonal events. They found that the size and the activ-
ity of the convective envelope played a dominant role in lowering
the isotopic composition of precipitation. This was supported by
the significant correlation between the isotopic composition of
local precipitation and the regional precipitation amount, and also
with the integrated precipitation amount of the air mass before ar-
rival at the measurement site.Inland troughs are also common over the Australian continent, par-
ticularly during the warmer months of the year. Frontal systems domi-
nate in the south, predominantly in winter, whilst low-pressure
systems are important for coastal regions.Multiple studies investigating
the impact of these synoptic weather systems, and of varying moisture
sources, on the isotopic composition of precipitation in SE Australia
have been undertaken in sites in Tasmania (Treble et al., 2005a; Barras
and Simmonds, 2008), Melbourne (Barras and Simmonds, 2009), Ade-
laide (Guan et al., 2013, 2009), at four sites in the Sydney region
(Hughes and Crawford, 2013; Crawford et al., 2013), at the Macquarie
Marshes (Crawford et al., 2017), and at the Snowy Mountains (Callow
et al., 2014).
Previous interpretations of the spatial characteristics of isotopes in
precipitation across the Australian continent by Liu et al. (2010) and
Terzer et al. (2013) were based on data from six coastal sites and only
one inland site (Alice Springs), as recorded in the Global Network for
Isotopes in Precipitation (GNIP) database (IAEA/WMO, 2017). The lim-
ited spatial distribution of data available at the time of these studies
meant that the range of climatic processes affecting rainfall isotopes
across the Australian continentwas not fully revealed. This paper builds
significantly on these earlier studies, through the presentation and in-
terpretation ofmuch-needed data from an expanded network of rainfall
collection sites across the Australian continent, and fills a gap in these
important records in the Southern Hemisphere. The expanded rainfall
sampling network developed for this study included the recommence-
ment of sampling at all seven of the original Australian GNIP sites, but
more importantly included the addition of seven inland continental
sites and one further coastal site to improve spatial distribution across
the Australian continent (see Fig. 1a and Table 1). Additional data
from this study offers improved spatial characterisation of isotopes in
rainfall across the Australian continent, as well as extending the GNIP
record a further 8–12 years in all stations, enabling characterisation of
seasonal trends and the environmental controls on isotopes in Austra-
lian rainfall.
This paper also develops and presents the first set of rainfall
isoscapes for Australia, allowing for the detailed prediction of the isoto-
pic signature in precipitation at any point on the Australian continent.
We demonstrate the predictive capability of these isoscapes and their
utility in informing groundwater studies in rainfall isotope data poor re-
gions, through the presentation of a number of case studies of ground-
water systems across NE Australia. Understanding the linkage
between rainfall and groundwater recharge processes in aquifers con-
taining groundwater with residence times that span several orders of
magnitude in timescale is vital for its management and sustainability.
This tool will help to better estimate recharge processes and therefore
incorporate more realistic predictions in groundwater models.
2. Site description and methods
2.1. Australian climate and weather
Approximately 70% of the Australian continent is classified as arid
(desert, average annual rainfall b250mm) or semi-arid (grasslands, av-
erage annual rainfall between 250 and 350 mm) and seven of the fifteen
rainfall collection sites in this study fall within these categories (Fig. 1
(a), Table 1). These are areas where rainfall is low Fig. 1(b) and unpre-
dictable, with considerable inter-annual variation.
Fig. 1(c–e) shows the typical weather patterns and related air mass
distributions that occur across the Australian continent during summer
and winter respectively. The monsoonal circulation and the formation
of a series of lowpressure cells to thenorth of Australia (Fig. 1(c)) brings
very warm, moist and unstable air to the north and north-west of Aus-
tralia,where it converges across a broad regionwith tropical continental
(Tc) and tropical maritime (Tm) air masses, shown as the intertropical
convergence zone (ITCZ) in Fig. 1(d). The rising air along this zone re-
sults in extensive convective activity and rainfall - known as the wet
632 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645
(a) (c) SUMMER (e) WINTER
(d) (f)
(b)
Fig. 1. Australian climate and sampling locations (a) major climate classes based on a modified Köppen classification (Stern et al., 2000). The black squares represent sites for which
historical data was available from the GNIP database. The green triangles represent the new sites in the expanded sampling network where no historical rainfall isotope data was
available. (b) Annual average rainfall (BOM) and groundwater sites. Typical air-mass distribution (c, e) and weather patterns (d, f) across Australia during summer (February (c, d))
and winter (July (e, f)).
(Source: modified after Sturman and Tapper, 2006).season. Tropical maritime air flows westward over the central part of
the continent, carried onshore by the trade winds. Orographic lifting
of this air along the Queensland coast leads to substantial rainfall. As it
moves inland it rapidly modifies to become very dry, hot and unstable
continental air.
Across southern Australia, as the high pressure systems migrate
from west to east, they bring a cycle of hot, dry weather followed by
cooler, moister southerly changes. The cold fronts associated with
troughs between these highs tend to be weaker in mid-summer.
In winter, the pattern of circulation features changes as themajor air
masses over Australia move north. The high pressure belt moves north
to lie over the cooler centre of Australia, and the highs move slowly
fromwest to east punctuated by cold fronts. A ‘baric ridge’ links the Pa-
cific and IndianOceanhighs. Thewesterlies bring cool,moist and cloudy
southern maritime (Sm) air to the southern part of the continent. This
region is dominated by the occurrence of cold fronts that develop out
of the intense low pressure cells in the Southern Ocean (Fig. 1e). Polar
air interactswith subtropical air, and a continuous cycle of frontal devel-
opment is embedded in the prevailing westerly airflow, resulting in
substantial precipitation (Sturman and Tapper, 2006). Snow fall in the
limited alpine areas occurs when modified polar maritime air (NPm)
is drawn from high southern latitudes behind a cold front. In northern
Australia, the low pressure cells that were a feature of summer have
moved further north along with the ITCZ. The south-east trade winds
cover much of the northern region, and the associated stable tropical
continental or tropical maritime air masses bringing cooler, drier
weather to the area (Fig. 1(f)).Intense low pressure systems are responsible for many of the severe
weather events that result in intense and long duration rainfall along
coastal areas of Australia: Tropical cyclones influence the northern
parts of the country (mostly in the warmer months); and East Coast
Lows occur along the east coast from southern Queensland to Victoria.
Large scale remote climate drivers, such as El Niño-Southern Oscillation
(ENSO), the IndianOcean dipole (IOD) and southern annularmode, also
have an impact on the rainfall variability over Australia (Risbey et al.,
2009).
The rainfall distribution across most of Australia (Fig. 1(b)) is
strongly seasonal in character with: a wet summer and dry winter re-
gime in the north (e.g. Darwin, Mt. Isa); a wet summer and relatively
dry winter in south-eastern Queensland and north-eastern New South
Wales (e.g. Brisbane, Charleville); fairly uniform rainfall in south-east-
ern Australia (e.g. Sydney, Melbourne, Wagga Wagga, Cobar); a wet
winter and dry summer regime in parts of the southern South Australia
and south-western Western Australia (e.g. Adelaide, Perth); and low
and erratic rainfall throughout much of the western and central inland
(e.g. Alice Springs, Mount Isa, Meekatharra, Woomera).
2.2. Sampling sites and history
This paper utilises 8–12 years of new precipitation stable isotope
data for 15 Australian sites as well as historical data from seven of
these sites available from the GNIP database (IAEA/WMO, 2017) (see
Fig. 1(a) and Table 1 – GNIP Site IDs marked with an asterisk). Sample
collection for six of the original seven GNIP sites commenced in 1962–
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 633
Table 1
Characteristics of the 15 Sampling Sites including the Australian Bureau of Meteorology (BOM) designated Site Name and Number, Data collection period and number data points (n), Latitude (decimal), Longitude (decimal), Elevation (m above sea
level), Annual Mean Precipitation, and Annual Mean Surface Air Temperature, Rainfall Climate Class and Modified Köppen Climate Classification (Stern et al., 2000). Site IDs marked with an asterisk (*) include stable isotope data currently available
from the GNIP database.
Full BOM Site Name BOM Site Data collection period n Latitude Longitude Elevation Distance from Precipitation Temperature (oC) Rainfall Climate Modified Köppen climate 
(Site ID) Number (oN) (oE) (m asl) coast (km)1 (mm) Class2 classification
Darwin Airport 1963–1966; 1973–1991; Equatorial savannah (dry 
14015 320 -12.4239 130.8925 30.4 5 1729.3 27.6 Summer Dominant
(Darwin*) 1995–2014 winter)
Eagle Farm Racecourse* / 40212 / 1962–1966; 1973–1982 / -27.4304 153.0672 7 10
Brisbane Regional Office* / 40214 / 1983–1991 / -27.48 153.03 38 15
466 1013.5 20.4 Summer No dry season
Brisbane Aero 40842 1995–2014 -27.3917 153.1292 4 4
(Brisbane)
Perth Airport / 1962–1966; 1973–1976 / -31.95 115.97 15.4 21
Perth CSIRO / 09021 (Perth 1983–1986; 1988–2000 / -31.95 115.78 16 3
348 775.1 18.3 Winter dominant Distinctly dry summer
Perth Airport Airport only) 2005–2014 -31.9275 115.9764 15.4 21
(Perth/Perth CSIRO*)                           
Adelaide Airport 1962–1967; 1972–1976; Distinctly dry (and hot) 
23034 219 -34.9524 138.5204 2 5 447.0 16.5 Winter
(Adelaide*)                        1983–1984; 2005–2014 summer
Cape Grim BAPS 1985–1991; 1994–95; 1997– No dry season (mild 
91245 274 -40.6828 144.6897 94 1 778.6 13.1 Winter
(Cape Grim*)                       2014 summer)
1962–1965; 1968–1969; 
Melbourne Airport 72 (23 to Port No dry season (warm 
86282 1973–1991; 1994–1998; 389 -37.6633 144.8339 113.4 542.2 14.6 Uniform
(Melbourne*) Philip Bay) summer)
2005–2014
Sydney Airport AMO No dry season (hot 
66037 2006–2014 95 -33.9411 151.1725 6 4 1085.1 17.8 Uniform
(Sydney)                      summer)
Wagga Wagga AMO No dry season (hot 
72150 2006–2014 99 -35.1583 147.4573 212 255 574.8 15.6 Uniform
(Wagga Wagga)                         summer)
Alice Springs Airport 1962–1970; 1972–1976; 
15590 204 -23.7951 133.889 546 895 284.9 21.0 Arid Hot (persistently dry)
(Alice Springs*)                   1983–1987; 2005–2014
Charleville Aero
44021 2006–2014 83 -26.4139 146.2558 301.6 554 497.7 20.8 Summer Hot (persistently dry)
(Charleville)                       
Cobar MO
48027 2006–2014 84 -31.484 145.8294 260 554 402.9 19.0 Uniform Hot (persistently dry)
(Cobar)                               
Meekatharra Airport
07045 2006–2014 74 -26.6136 118.5372 517 430 236.6 22.4 Arid Hot (summer drought)
(Meekatharra)  
Mildura Airport
76031 2006–2014 92 -34.2358 142.0867 50 307 292.5 17.1 Winter Warm (persistently dry)
(Mildura)                         
Mount Isa Aero
29127 2006–2014 62 -20.6778 139.4875 340.3 334 469.8 24.6 Summer Dominant Hot (winter drought)
(Mount Isa)                         
Woomera Aerodrome 286 (178 to 
16001 2007–20113 46 -31.1558 136.8054 166.6 185.9 19.2 Arid Hot (persistently dry)
(Woomera)                       Spencer Gulf)
1Distance from coast given for current sampling location only.
2Summer dominant =marked wet summer and dry winter, Summer=wet summer and low winter rainfall, Uniform= uniform rainfall, Winter=wet winter and low summer rainfall, Winter dominant =marked wet winter and dry summer,
Arid= low rainfall.
3Four additional samples only from 2006, 2012, 2013.
A
D E S E R T G R A S S L A N D T E M P E R A T E S U B - T R O P I C A L T
634 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 63563, with collection commencing in 1979 at Cape Grim. For the purposes
of statistics used in this study, stable isotope data from Cape Grim that
was collected prior to 1985 has not been used as the correspondingme-
teorological data (other than precipitation amount) is not available. As
can be seen in Table 1, the data collection period was not continuous,
with only three sites, of the original seven sites, still being actively sam-
pled by 2005. Data after December 2002 was not available through the
GNIP database at the time of writing this paper. Missing stable isotope
data for monthly rainfall samples collected from 2003 to 2005, as well
as for some samples collected during 2001–2, were obtained directly
from the CSIRO Land and Water Isotope Laboratory research team,
who had been analysing the rainfall samples since 1982. In 2005, a col-
laborative arrangement between ANSTO, CSIRO, and the Australian Bu-
reau ofMeteorology (BOM)agreed to re-instate sampling at those of the
original seven sites which had ceased operating. In 2006, ANSTO insti-
gated an expansion of the Australian rainfall sampling network to a
total of 15 sites to include more inland stations (see Fig. 1(a)).
Historical monthly δ2H and δ18O data, as well as corresponding
monthly temperature, precipitation amount and vapour pressure data
for eight of the Australian stations were obtained from the GNIP data-
base (accessible at http://www.iaea.org/water). Meteorological data
gaps have beenfilled,when available, using data from theAustralian cli-
mate data archive provided by the Australian Bureau of Meteorology
Data Service (see Table 1). All rainfall samples from 2005 onwards
were collected and analysed as outlined in the following section.
2.3. Sample collection method and analysis
Rainfall samples collected since inception of the IAEA/WMO pro-
gram in the 1960s were obtained using a composite sampler comprised
of a funnel with a rubber bung in a large glass or steel collector. One ex-
ceptionwas the CapeGrim sitewhere first an ERNI rainfall collector, and
later an EIGENBRODT rainfall collector was used; both had a lid that au-
tomatically operated at the onset of rainfall. Whilst the authors cannot
be certain of the rainfall sample collection methods for stable isotope
analyses prior to 2005, there is no evidence that any technique to pre-
vent evaporation of the sample (e.g. paraffin oil) other than shielding
the collection vessel from sunlight and wind was undertaken.
All samples from 2005 onward (except Darwin) were collected as
a manual monthly composite of daily rain gauge samples, following
the technical procedure recommended for GNIP sampling at http://
www-naweb.iaea.org/napc/ih/documents/userupdate/sampling.
pdf. This reduces potential evaporative enrichment and vapour ex-
change compared with the previous technique. However, of the 15
sites Darwin continued to use the previous method. Darwin rainfall
is highly seasonal with high monthly rainfall amounts during the
monsoon season. This would reduce the impact of evaporative en-
richment at this location. A comparison of the weighted and arith-
metic averages between the current dataset, and the data from
1962 to 2002 (Liu et al., 2010) shows no systematic differences in
d-excess that would indicate that evaporation has significantly af-
fected either the historical or recent datasets.
Isotope analyses for samples collected from 2003 to October 2010
were undertaken using Isotope Ratio Mass Spectrometry at the
CSIRO Land and Water Isotope Lab (Adelaide) (reported accuracy
of ±1.0, ±0.15‰ for δ2H, δ18O) or Alberta Innovates Technology
Futures Isotope Hydrology and Geochemistry Lab (reported accuracy
of ±1.0, ±0.2‰ for δ2H, δ18O). All samples collected from October
2010 onwards were analysed at the ANSTO Environmental Isotope
Laboratory using a Cavity Ring-Down Spectroscopy method on aFig. 2. The seasonal variability of rainfall isotope data (weightedmonthlymeans for δ18O and co
the 15 stations currently in operation across Australia. For rainfall and temperature, the mon
collected at a site [dataset] and the long-term monthly climate averages representing all availaPicarro L2120-I Water Analyser (reported accuracy of ±1.0, ±
0.2‰). All isotope results are reported as per mil (‰) deviations
from the international standard, V-SMOW (Vienna Standard Mean
Ocean Water). The analytical laboratories for earlier data are listed
in the GNIP database.
2.4. Isotopic data quality
The authors have chosen not to “clean up” the isotopic data by fol-
lowing a common method of quality control, which includes removing
outliers that lie outside the 2σ or 3σ band of the regression between
δ18O and δ2H (IAEA, 1992). The data from numerous event based pre-
cipitation sampling studies have shown that the variability in isotope
values can be large. The “extreme” values identified as lying outside
the 3σ band of the regression linemay in fact signify real episodes of un-
usual non-equilibrium conditions that may warrant further investiga-
tion of the climate/synoptic conditions producing them. For instance,
many of the most depleted values in this dataset result from extreme
low pressure events, some of which lie outside the 3σ band because of
their abnormally low d (e.g.WaggaWagga, ex Tropical Cyclone Oswald,
January 2013; Darwin, Tropical Cyclone Gretel, April 1985). Some au-
thors have also chosen to remove isotope data from low rainfall months
(e.g. b2–5 mm/month, Hughes and Crawford, 2013) or with low d
values (e.g. b3‰, Harvey (2001)) from their statistical treatments, as
these samples may have been influenced by evaporation whilst sitting
in the collector. However, as many inland sample sites in this study
are located in arid and semi-arid areas, the low d values observed may
in fact be because of sub-cloud evaporation of the rainfall, and not be-
cause of sampling errors. We have therefore chosen to include isotope
data from all rainfall samples which have been collected, aside from ap-
plying a cut-off of precipitation amounts of b2 mm (due to the high fre-
quency of very low d values associatedwith these samples; for example,
at Woomera five such samples had a d of −7.8‰ or less).
2.5. Meteoric water line methods
Local meteoric water lines (LMWL) were calculated using two re-
gression models: ordinary least squares regression (OLSR) and precipi-
tation amount weighted least squares reduced major axis regression
(PWRMA). Precipitation weighted methods for calculating a LMWL re-
duce the influence of small rainfall events, which have been potentially
affected by evaporative enrichment, in the calculation of the slope of the
LMWL (Hughes and Crawford, 2012; Crawford et al., 2014) and reduce
the need for arbitrary cut-offs. This is most applicable in hydrological
studies where these low rainfall events are insignificant in terms of run-
off and recharge events, andwhere it ismore appropriate to use a LMWL
weighted more towards high rainfall episodes. The details of the OLSR
and PWRMA models are outlined in Crawford et al. (2014); PWRMA
was chosen because a RMA regression is more appropriate when two
variables are related by underlying physical processes as it optimises
the fit to both δ2H and δ18O (IAEA, 1992).
3. Results and discussion
3.1. Seasonal and spatial variation of the δ2H, δ18O and d data
The long-termmonthly average temperature and rainfall (calculated
on variable length datasets; ranging from 28 years for Cape Grim up to
71 years for Alice Springs) and the corresponding monthly averages
for the sampling period are presented in Fig. 2. Average rainfall andrresponding deuterium excess, dP), rainfall amount (bars) and temperature (diamonds) for
thly averages representing the same monthly intervals for which rainfall samples were
ble temperature and rainfall measurements for a site [climate] are shown.
636 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645temperature calculated for months with corresponding isotopic data
(dataset) generally compare closely to long-term climate averages (cli-
mate) indicating that the sampling period is climatically representative.
However some notable differences between the monthly rainfall aver-
ages for the dataset and long-term climate value do exist, e.g. February
at Brisbane;March andDecember at Charleville; June atMt. Isa; January,
May and June at Sydney; March in Wagga Wagga and February at
Woomera. Those months in which the dataset average rainfall was
higher than the long-term climate average were mostly associated
with the capture within the sampling dataset of extreme events
resulting from East Coast Lows, trough systems and residual moisture
from low-pressure systems including ex-Tropical Cyclone Yasi (2011).
Other months in which the dataset average rainfall amounts were con-
sistently lower than the long-term averagemay reflect an absence of in-
frequent extreme events during the sampling period. Supplementary
Fig. S1 also gives an indication of how climatically representative the
sampling period in this study was. Fig. S1 shows the total accumulated
precipitation sampled for each month at each site and the correspond-
ing interquartile range of δ18O values. In general, the range in monthly
isotopic values is greater than the range for the inland sites. The var-
iability in isotopic mean/median between months is generally
greater for those sites with a smaller number of data points indicat-
ing that a longer dataset may be required to consider the statistical
monthly variation.
Fig. 2 illustrates the seasonal patterns in the stable isotopes ratios
δ2H, δ18O in Australian precipitation. A weak seasonal variation in the
isotopic composition is observed with winter minimum and spring-
summer maximum in the coastal temperate sites of Adelaide, Brisbane,
Cape Grim, Melbourne and Sydney. Inland sites (Alice Springs,
Charleville, Cobar, Mildura, Mt. Isa,WaggaWagga, andWoomera) com-
monly exhibit a springmaximumand lower values in summer andwin-
ter. A springmaximumwas also found for four sites in the Sydney Basin
(Hughes and Crawford, 2013) which was attributed to (1) smaller pre-
cipitation amounts in spring, and sub-cloud evaporation; and (2) first-
stage rainout. Both of these processes result in more isotopically
enriched rainfall and would be considered valid for the inland sites in
this study. These sites generally experience dry winters and the humid-
ity tends to be lower in winter to springwhichwould also increase sub-
cloud evaporation. Darwin, which is subject tomonsoonal rain, has a bi-
modal distribution with low values during thewet season fromDecem-
ber to April and high values during the dry season. Feng et al. (2009)
found a local isotopic minimum near the equator which coincides
with the movement of the Intertropical Convergence Zone (ITCZ), also
identified by Matsui et al. (1983) at Belém (Brazil) and Edirisinghe et
al. (2017) in Sri Lanka. We see strong evidence for this not only in Dar-
win but in other northern Australian sites, such as Mount Isa and Alice
Springs, which are influenced by the monsoon. In mid-summer the
southward shift in the ITCZ, which results from the convergence of
equatorial maritime with tropical continental and tropical maritime
air, causes air to rise producing significant precipitation over northern
Australia (Sturman and Tapper, 2006).
As expected, the d is higher in the austral winter months (June–
August) than in the summer months (December–February) for all
sites except Mt. Isa, where the winter record is dominated by a single
depleted extreme event in June 2007 which accounts for 75% of all
winter rainfall sampled in this study. Isotopically enriched rainfall
with low d was recorded at Mt. Isa in August to October, which cor-
responds to small rainfall events, suggesting re-evaporation of rain-
drops below the cloud. The higher d at other sites in winter may be
attributed to cold air masses with low relative humidity travelling
over the ocean, thus resulting in higher kinetic fractionation during
evaporation at the moisture source (Liotta et al., 2006; Pfahl and
Sodemann, 2014). Comparing weighted averages for all samples
with those with precipitation ≥5 mm it can be noted that whilst the
subset with the small rainfall amounts removed has a marginally
more depleted average, this difference is negligible for the coastalsites and below the reported analytical error for the inland sites,
and the difference in d is always b0.2‰.
Larger d valueswere seen in Sydney,WaggaWagga, Cobar and Perth
in particular. To examine if this was as the result of the shorter time se-
ries for these sites, precipitationweighted values for the shorter periods
(i.e. for years 2007–2014) were also calculated for the long-term sites
(Supplementary Table S1). A higher d resulted for the years from 2007
for a number of the long term sites, e.g. in Brisbane, but excluding
Perth, when using the entire data set the precipitation weighted dp
was 13.5‰, whereas when using data from 2007 onwards the dp
was 14.2‰ (comparable to Wagga Wagga), highlighting the value
of longer time series. Carrying out a trend analysis found a slight in-
creasing trend in δ18O over the period of record at Brisbane
(0.0075‰ per year) and Adelaide (0.008‰ per year) and also in d
at Brisbane (0.01‰ per year) but no significant trends at the other
sites. Other than Darwin and Melbourne the long term precipitation
weighted average values of δ18O and δ2H are more enriched in the
current longer data set than that in Liu et al. (2010). In particular,
the 2007–2014 period had significantly more depleted precipitation
at Darwin because the historical dataset had a number of very large
(N700 mm/month) rainfall events that were more enriched than
the general precipitation amount effect would predict. Brisbane
also exhibits a difference with a reduction in the amount of depleted
rainfall months in the 2007–2014 period.
Annual, seasonal and monthly arithmetic and weighted average
values of δ2H, δ18O and d are listed in Supplementary Table S1 and the
monthly data plotted in Supplementary Fig. S2. The weighted means
for all sites across the Australian continent are more depleted than the
arithmetic mean, reflecting the seasonality in rainfall amounts and iso-
topic values. In addition, the weighted mean dp values are higher than
the arithmetic mean d values for inland sites in particular, suggesting
that sub-cloud evaporation is important. Guan et al. (2013) found that
where sub-cloud processes have an impact, the precipitation-weighted
mean is lower than the arithmetic mean and the precipitationweighted
dp value is higher than the arithmetic mean. This is seen at most sites
and is particularly evident at the arid inland site, Alice Springs (Supple-
mentary Table S1 and Fig. S2).
Themost isotopically depletedmonthly rainfall samples in the entire
dataset occurred at Mount Isa. In order to understand processes leading
to this depletion, back trajectories using HYSPLIT (Draxler and Rolph,
2003) were generated for some of the months with the most isotopi-
cally depleted precipitation. In these cases it was found that substantial
precipitation had occurred prior to the sampling site (using theHYSPLIT
generated precipitation amount at each back trajectory location before
the sampling site), thus indicating isotopic depletion due to the prior
rainout. Heavy and prolonged rainfall occasionally falls at Mount Isa
with the passage of ex-tropical cyclones (e.g. cyclone Ellie January–Feb-
ruary 2009, forwhich 18O depleted rainfallwas sampled). Themean sta-
ble isotope ratios of rainfall from tropical cyclones were found by
Lawrence and Gedzelman (1996) to be distinctly lower than those
of other tropical and summer rainfall events, which can be attributed
to the high condensation efficiency of tropical cyclones and storms.
This has a definite effect on precipitation at both Darwin and
Mount Isa, and is occasionally seen at other sites such as Brisbane,
Charleville and Meekatharra, and further south. The isotopic evolu-
tion of precipitation from tropical cyclone Ita (April 2014; Trinity
Beach P = 231 mm, δ18O = −10.2‰), in north-eastern Australia,
has been detailed in Munksgaard et al. (2015) and isotopic values as-
sociated with cyclones in north-west Australia are also reported to
be significantly depleted (Skrzypek et al., 2018). However, whilst
some of the most depleted rainfall in this dataset is due to tropical
cyclones and low-pressure systems, not all of these events result in
depleted rainfall. The weighted average composition for 15 individ-
ual months when tropical cyclones affected Darwin was more
depleted (precipitation weighted average δ2H of −40.9 and δ18O of
−6.4) than the average of all data, however, five of those months
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 637were more enriched than the average. This is most likely due to the
observed behaviour of the isotopic composition of precipitation
from cyclones, which is variable depending on the distance from
the eye of the cyclone (e.g. Munksgaard et al., 2015).
3.2. Local meteoric water lines
The relationship between δ18O and δ2H, known as the meteoric
water line, provides a useful frame of reference for comparing different
sites and regions, as well as for hydrological and groundwater studies
where rainfall is a significant end member. Both ordinary least squares
(OLSR) and precipitation weighted reduced major axis (PWRMA) re-
gression models (Crawford et al., 2014) were used to determine Local
MeteoricWater Lines (LMWL) for the Australian sites using all available
monthly data from the 15 stations (Fig. 3; Supplementary Table S2).
There is a significant increase in the slope and intercept of the LMWL
calculated using PWRMA relative to that of the OLSR. This reflects the
removal of a bias in the calculation of the LMWL towards the small rain-
fall events which may have experienced secondary evaporation effects
during rainout, resulting in lower d and more enriched isotopic values.
These events occur most frequently in the arid and semi-arid interior
of Australia (e.g. Macquarie Marshes, NW-NSW; Crawford et al., 2017;
where the PWRMA slope was 7.2, still lower than the slope of 8 for
the Global MWL) where higher temperatures and low humidity are ex-
perienced for much of the year, as well as the Mediterranean climate
zones in South and Western Australia which experience hot dry sum-
mers (Hughes and Crawford, 2012). Nine of the fifteen sites in this
study lie in these zones.
In order to compare seasonal variations, the LMWLswere also deter-
mined using both regressionmethods for the summer (DJF) andwinter
months (JJA). Generally, the slopes are greater in winter for the south-
ern coastal and inland sites and greater in summer for the northern
and eastern coastal sites, broadly reflecting the seasonal rainfall domi-
nance. However, it also observed that there was a group of sites with
significantly less difference between summer and winter that appeared
to correlate with shorter datasets used to calculate the LMWLs. We ini-
tially considered that this may be a statistical artefact whereby extreme
rainfall events may bias the calculations for the relatively short seasonal
datasets (e.g. Mount Isa andWoomera had data from only 6 and 8 win-
ter samples, respectively). To test this theory, LMWLs were developed
for Melbourne and Perth using only data from the time period corre-
sponding to the more recent sites (i.e. 2006–2014). It was found that
the differences in the slopes of these seasonal LMWLs for these abridged
datasets did not increase, and therefore we do not believe the observed
differences were due to statistical artefacts. As most of the sites with
longer datasets are located on the coast, we also considered the possibil-
ity that there may be a moderation in the difference between seasonal
LMWLs due to the close proximity of the sites to the dominantmoisture
source for the rainfall. However, this is unlikely as a significant differ-
ence was also observed at Sydney, which is also located on the coast.
Hence, we consider that the variability between the seasonal LMWLs
is not simply attributable to a seasonal effect or geographical location,
but is also due to other processes such as the different weather systems
contributing to rainfall at each of the sample sites and below cloud
evaporation in hot and dry locations. For example, the smaller slope of
the LMWL observed at Alice Springs in summer, could be attributed to
below cloud evaporation (e.g. Harvey and Welker, 2000). The larger
slope in the observed LMWL at Darwin during the summer season is
consistent with other sites affected by monsoonal rainfall (e.g. Kumar
et al., 2010).
The use of our measured monthly rainfall data to construct an Aus-
tralianMWL could be of considerable value for use in comparing rainfall
isotopic data in Australia with other continents, regions or islands, or as
a frame of reference for other Australian precipitation data. However,
the weighting of the stations to construct such aMWL needs to be care-
fully considered because the distribution of our study sites does not fullyrepresent the range of climate and geography in Australia, and there are
large differences in rainfall amount between sites. The use of precipita-
tion weighted regressions, in this case, has the potential to introduce an
undesirable bias towards high rainfall coastal sites and extreme events
which may not reflect the precipitation over the majority of the inland
areas of the continent (Crawford et al., 2014). Use of all available
monthly data could exacerbate this because it is mainly the coastal
sites that have long data records and a larger number of samples per
year than the drier inland sites. An Australian rainfall isoscape was
also developed in this study (Section 3.3.5) in order to estimate the spa-
tial distribution of isotope values for regions where no sample data is
available. The gridded precipitation δ18O and δ2H values which were
predicted through the development of this isoscape can also be used
to develop a national MWL – the advantage of this approach is that it
can result in a more geographically (non-weighted) or hydrologically
(precipitation weighted) representative MWL for the entire country or
any sub- region.
Using the isoscape gridded precipitation data, the non-weighted
RMA derived Australian MWL was calculated to be δ2H = 8.6δ18O +
14.4 and the precipitation weighted RMA MWL was δ2H= 8.6δ18O +
15.3. As indicated above, we recommend the use of these MWLs as
most representative for the Australian continent. For comparison, we
also present Australian MWLs calculated using the available sample
data: (1) a non-weighted RMA regression on the weighted average an-
nual values for all 15 sites (calculated from monthly samples,) gives a
MWL of δ2H = 8.3δ18O + 14.1, (2) a precipitation weighted RMA re-
gression on all measured monthly rainfall data (n = 2855) gives a
MWL of δ2H = 7.9δ18O + 12.3, and (3), RMA regression on precipita-
tion weighted monthly average values gives a MWL of δ2H= 7.2δ18O
+ 7.6 (parameters for all regressions are given in Supplementary
Table S2). The difference between each of these calculatedMWLs is sig-
nificant, and highlights the need to take care in the development of re-
gional MWLs from a spatially irregular sampling network.
3.3. Environmental controls on stable isotopes in Australian rainfall
3.3.1. Temperature effect
As decreasing temperature drives the rainout process, a strong cor-
relation between temperature and isotopes in precipitation should
occur, i.e. where a temperature gradient exists, there will be a gradient
in the δ18O and δ2H signatures. Dansgaard (1964) identified a global lin-
ear relationship between surface air temperatures and mean annual
δ18O in precipitation as
δ18O ¼ 0:695Tannual−13:6‰
However, at the regional to local scale, the T-δ18O relationship may
not hold true because of the dominance of continental or rainout effects,
the impact of sub-cloud evaporation, the seasonality of rainfall or phys-
iographic variations (Clark and Fritz, 1997). When the T-δ18O relation-
ship for each of the individual sample locations is calculated, the
slopes are all positive but much lower than the above global slope.
The r2 values are also generally low, when either the monthly sampled
data is used or annual average values are used in the calculations (Sup-
plementary Table S3). For the Australian continent, the linear relation-
ship between mean surface air temperature and mean annual δ18O in
precipitation (r2 = 0.23) is
δ18O ¼ −0:115Tannual−2:65‰
The slope of this relationship is influenced by the continental effect
and the fact that different processes dominate at the different sites, e.
g. at Alice Springs, Mt. Isa and Darwin, the amount effect has a higher
r2 value than the temperature effect (Supplementary Table S3;
discussed in the next section).
638 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 639
Fig. 4. Amount weighted monthly and annual mean δ18O versus mean air surface
temperature. Groupings are: Southern Inland (Melbourne, Wagga Wagga, Cobar,
Mildura and Woomera), Coastal (Adelaide, Sydney, Brisbane and Perth), Northern
Inland (Alice Springs, Charleville and Meekatharra), Tropical (Darwin and Mount Isa).
Individual monthly data for the sites in each group are represented by dots of the same
colour.On closer examination of the annual mean (temperature, δ18O) rela-
tionship in Fig. 4 we see three main groups, from left to right: (1) an
enriched value at the lowest temperature (Cape Grim, which is located
the furthest south and receives first stage rainout from frontal systems
passing across the Southern Ocean); (2) a group of values (southern in-
land and coastal sites) in which a positive relationship between δ18O
and temperature exists; (3) followed by a group of values with almost
very little change in δ18O with temperature, which consists of the trop-
ical sites (Darwin andMt. Isa, for which isotopes are depleted due to the
amount effect) as well as the northern inland sites of Charleville, Alice
Springs and Meekatharra, for which there is depletion due to the conti-
nental effect. If the first and last sets are removed a better relationship
exists between temperature and δ18O for the southern inland and
coastal sites, with a regression line of δ18O= 0.259 T− 9.07‰, and an
r2 of 0.47.3.3.2. Precipitation amount effect
The initial liquid phase of rainfall is enriched in the heavy isotopes
when compared to rainfall later in the rain event. As a consequence,
the rainfall becomes more depleted (lighter) in heavy isotopes as the
rain continues to fall, and this is known as the “amount effect” or “rain-
out effect” (Dansgaard, 1964). Risi et al. (2008) identify two processes
that explain the “amount effect” in convective systems, in the tropics:
the re-evaporation of falling rain and diffusive exchangewith surround-
ing vapour; and the recycling of the sub-cloud layer feeding the convec-
tive system.
A negative slope is seen between δ18O and precipitation amount at
all sites with a higher r2 value than that of local surface temperature
(Supplementary Table S3). In particular, the correlation is significant
for Sydney, Alice Springs, Mildura, Mt. Isa and Darwin. DansgaardFig. 3. Local meteoric water lines for each of the fifteen collection sites, calculated using all data
LMWLs are shown in blue (precipitation weighted reduced major axis regression, PWRMA, eq
Parameters for these LMWLs are listed in Supplementary Table S2. The precipitation weighted
Table S1).(1964) found that this effect is seen throughout the year inmost tropical
stations and in the summer at mid-latitudes. This is supported by our
findings of a stronger correlation between δ18O and precipitation in
summer at a large number of the sites.
The partial correlation coefficients presented in Supplementary
Table S3 are an indication of the covariance between the monthly rain-
fall δ18O values and the monthly rainfall amount (P; when the effect of
temperature has been removed) and average monthly surface air tem-
perature (T, when the effect of rainfall amount has been removed).
There is little difference between the correlation coefficients between
the raw and deseasonalised data, which indicates that the seasonality
of temperature and rainfall have little effect on the isotopic composition.
The exception is the northern-most sites affected by tropical cyclones
and the monsoon (Darwin, Mt. Isa, Meekatharra) which do show a
lower correlation for precipitation when temperature effects are
removed.
3.3.3. Relationship of δ18O with vapour pressure (ea) and relative humidity
(RH)
The relative humidity at a site has an impact on the sub-cloud evap-
orationwhich results in isotopicallymore enriched remainingprecipita-
tion (Dansgaard, 1964).We see a negative relationship between relative
humidity and precipitation δ18O values consistentwith this, particularly
for the inland sites, where the RH-δ18O correlation was generally better
than that for temperature or precipitation (Supplementary Table S3).
For the relationship between δ18O and vapour pressure the best r2
values were those for Mount Isa and Alice Springs. This correlation
was negative, implying that isotopically more depleted δ18O values
were seen under higher vapour pressure; however, the correlation for
the other sites was low.
3.3.4. Latitude, altitude and continental effects
A broad latitudinal gradient is observed across the continent (Fig. 5
(a)), with a general trend for increasing enrichment poleward (δ18O=
−0.066 × latitude − 6.84), rather than the opposite observed in the
global data-set for mid-latitudes (e.g. Rozanski et al., 1993). The ob-
served pattern is similar to that presented by Feng et al. (2009; where
a positive trend in δ18O was seen from the equator to about 30° north
and south and then a rapid decrease was observed) and reflects the
dominance of themonsoonal (Zwart et al., 2016) and tropical cyclone re-
gimes (Munksgaard et al., 2015; Skrzypek et al., 2018) to the north and
the diverse weather types and oceanic moisture sources impacting the
southern half of the continent.
The altitude range of the sites is small, with the lowest altitude of 2
m asl and the highest at 546 m asl. Even so, a trend is seen in the δ18O
values, with more depleted values at the higher altitudes (δ18O =
−0.0026 altitude −4.4‰; r2 of 0.28; Fig. 5(b)), whereas there is no
clear trend in d. An altitude effect is reported by Callow et al. (2014)
and Tadros et al. (2016) for the Snowy Mountains (950–2045 m asl),
and Hughes and Crawford (2013) for the Sydney Basin (152–1178 m
asl), with precipitation more depleted than the corresponding lowland
sites from this network (Wagga Wagga and Sydney). This indicates
that the expanded GNIP network is less effective for defining the isoto-
pic composition of rainfall along theGreat Dividing Range along the east
coast of Australia than it is for coastal and inland Australia. This is ex-
plained further in Section 3.3.5 where the Australian isoscapes are de-
veloped and discussed.
Landmasses force rainout from vapourmasses as they move inland,
so that the residual vapour and subsequent rainfall at the centre of a(circles). The dashed line represents the GMWL (δ2H= 8∙δ18O+ 10; Craig, 1961), and the
uation in bottom right hand corner) and green (ordinary least squares regression, OLSR).
dp, δ18O, δ2H, and sample number n are shown in top left hand corner (Supplementary
640 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645
Fig. 5. The distribution of the weighted annual δ18O values in precipitation with (a) latitude (p-value = 0.05), (b) altitude (p-value = 0.04), and (c) distance from the coast
(p-value = 0.13).large land mass or continent, such as Australia, becomes progressively
more depleted in heavy isotopes – the “continental effect”. However,
this effect may be counteracted by the contribution of isotopically
enriched recycled continental moisture back into the atmosphere
from evaporated soil and surface water. Furthermore, sub-cloud evapo-
ration also leads to isotopically more enriched precipitation, which can
also counter-act the continental effect; however, this has a lower impact
during large rainfall events. To examine if a continental effect exists for
the isotopic composition of Australian precipitation a linear regression
was undertaken for the δ18O with distance from the coast (Table 1).
The resulting regression line was δ18O = −0.0011 × distance −
4.54‰, with an r2 value of 0.16 (Fig. 5(c)). The correlation is low,
which may be because there are a range of climatic zones across the
continent. Not all sites receive the majority of their rainfall along a tra-
jectory perpendicular to the coast andmany sites, particularly in Eastern
Australia, receive significant rainfall from thenearest coastline aswell as
along continental trajectories 1000 km to the west. Along a transect
from Adelaide-Mildura-Wagga Wagga we see depletion (presumably
due to rainout), with precipitation weighted δ18O values of −4.35,
−5.22 and −5.39‰, respectively. Undertaking a regression fit of δ18O
with distance from Adelaide resulted in a regression line of δ18O =
−0.0032 × distance − 4.34‰, which is very close to the regression
based on all data, but with a much improved r2 value of 0.9 due to the
removal of variability due to other factors (such as varying moisture
sources and trajectories).
3.3.5. Development of an Australian isoscape
Interpolation of a sparse spatial dataset such as presented in this
paper is difficult. Simple interpolation methods such as kriging cannot
account for the climatic and geographic factors that influence precipita-
tion isotope composition. Interpolation predictionmodels incorporating
first latitude and altitude (Bowen and Wilkinson, 2002), distance from
the coast (Bowen and Revenaugh, 2003), then adding meteorological
variables (e.g. Lykoudis and Argiriou, 2007) and, most recently, climatic
zones (Terzer et al., 2013) are believed to be an improvement on simple
interpolation.
The Regionalised Cluster-Based Water Isotope Prediction (RCWIP)
gridded precipitation δ18O and δ2H (Terzer et al., 2013; IAEA, 2014)
overestimates both δ18O and δ2H for most months at the 15 sites in
this dataset, except for CapeGrim. The correlation between gridded out-
puts and annual weighted averages was moderate to strong for the
coastal sites with r2 ranging from 0.41 to 0.80 (Supplementary Table
S4), with the exception of Perth and most of the inland sites. Generally,
RCWIP predicts the winter δ18O and δ2H well at most Australian sites,
but performs very poorly in summer, when it overestimates both δ18O
and δ2H by more than the interpolation uncertainty for many months
at all sites (excluding Cape Grim, where it is underestimated) and up
to ~11.8‰ for δ18O and 92‰ for δ2H for January. RCWIP uniformlypredicts awinterminimumand summermaximum,whereasmost Aus-
tralian sites have their most enriched rainfall in spring-early summer
and many have their most depleted rainfall in late summer-autumn.
In zones with winter dominant or uniform rainfall distribution, a rea-
sonable annual weighted average match is seen despite RCWIP's inabil-
ity to match the observed isotopic seasonality. Where the rainfall
climate is arid or summer dominated RCWIP performs more poorly
and in the northern half of Australia, there may be an improvement to
be made by having additional sites to define the regression parameters
for this climate zone. To achieve this we used additional rainfall isotope
data from sites shown in Fig. 6a (Cartwright and Morgenstern, 2016;
Cook et al., 2001; Crawford et al., 2017; Crosbie et al., 2012; Goede et
al., 1982, 1986, 1990; Hughes and Crawford, 2013; Munksgaard and
Zwart, 2017; Tadros et al., 2016; Treble et al., 2005b; Skrzypek et al.,
2018; Stephen Lewis, James Cook University, pers. comm 17 June
2015; Supplementary Table S5b).
To generate isoscapeswe used relationships across the continent be-
tween geographic parameters and the precipitation weighted average
δ18O, δ2H and dp values from each site. Aside from using the geographic
variables as predictors for isotopic composition, we investigated if long-
term average precipitation, temperature, relative humidity and vapour
pressure at the sites could also be used as predictor variables in the de-
velopment of an isoscape. We found that relative humidity, vapour
pressure and precipitation amount were poor predictors of rainfall iso-
topes on a continental basis. A significant correlation was found when
temperature was used, by itself; however, when latitude and altitude
were also included in the regression the marginal contribution of tem-
perature was low, hence temperature was not used in the isoscape
development.
Due to the large number of climatic zones andmoisture sources over
the Australian continent and the small number of sampling sites for a
number of these climatic types, wewere unable to generate an isoscape
using themethod presented by Terzer et al. (2013). However, by group-
ing the available sites into four coastal regions (east, west, south and
north) and an inland group (Supplementary Table S5) we were able
to obtain regression relationships of δ18O with a combination of predic-
tor variables; latitude, altitude and distance from the coast. The regres-
sion equations had the following form (following from Bowen and
Revenaugh, 2003, with the addition of distance from the coast):
y ¼ aðLatÞ2 þ bjLatj þ cAlt þ dDistance ð1Þ
where y was δ18O, δ2H or dp and Distance was the distance from the
nearest coast. For the north, west and south sites, latitude, altitude
and distance from the coast were used. For the east and inland groups
only latitude and altitude were used. For the east sites, the distance
from the coast was not significant and did not increase the r2 of the fit.
Whereas, only five sites were available for the inland group and
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 641
Fig. 6. a) Digital elevation model and location of all sites used in isoscape development. Annual isoscapes (‰) for b) δ18O; c) δ2H; and d) dp. GNIP sites are shown as red circles and
additional sites as white squares.generating a regressionwithout distance from the coast resulted in an r2
of 0.99. If more parameters were to be resolved in the regression fit we
found that this overfitting of the data presented some problems during
the use of the regression model for prediction. The determined param-
eters of the regression fits are presented in Supplementary Table S5.
Once the equations for each region were determined, the Australian
region was subdivided in to grid cells of 10′ and a value of the isotopic
composition for each grid cell was estimated using the regression equa-
tions. To overcome a subjective choice of defining a boundary between
the regions (for which a different regression fit was determined), a dis-
tance weighed procedure was used (similar to that used by Bowen and
Revenaugh, 2003, for the interpolation of the error). Basically, the esti-
mated isotopic value was a weighted sum from using each of the five
equations with the local values of the predictor variables. The weight
was the negative exponential of distance (in degrees) between the loca-
tion of the grid cell and each of the measurement sites, i.e.
Pn y e−ðDxi =BÞ
p ¼ i¼1 ix P x ð2Þn
i¼1 e
−ðDi =BÞ
where px is the predicted values of the isotopic composition in cell x, n is
the number of observation sites, yi is the regression equation associatedwith site i (but using the values for grid cell x for the predictor vari-
ables), Dxi is the distance between the grid cell x and the measurement
site i, and B is a smoothing parameter. The estimated isotopic value of
grid cell xwas then determined as the sum of px in Eq. (2) and an inter-
polated error term (errx):
Pn x
¼ err
−ðD
ie i
=BÞ
errx ¼ i 1P ð3Þn −ðDxe i =BÞi¼1
where erri is the difference between the measured value at site i and
that estimated using Eq. (2). The value of B in Eqs. (2) and (3) was esti-
mated by removing the measurement sites, one at a time and then
minimising the difference between the estimated and measured values
(resulting in a value of 1.45 for B). This was confirmed by a Moran's
(1948) test where correlations between the errors and the spatial loca-
tions existed, with a p-value of 0.07 for errors in δ18O and a p-value of
0.09 for errors in δ2H.
Once the isoscapes were developed a comparison was made be-
tween the estimated d using two methods; (1) using the measured d
values and developing an isoscape of d and (2) using the isoscape de-
rived values of δ18O and δ2H and using the relationship d = δ2H −
642 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–6458δ18O. The maximum difference between the d as calculated from the
isoscape and the d as calculated from δ18O and δ2H was 0.01%.
As in Terzer et al. (2013), jackknifing (Wu, 1986) was used to deter-
mine the robustness of themodel results. In this approach, all siteswere
removed one at a time and the values re-estimated over the Australian
region. Then the mean and standard deviations of the estimates at each
grid cell were determined. However, two approaches were imple-
mented. In the first approach (regression jackknife), when a site was re-
moved the regression models were re-determined; in the second
approach (interpolation jackknife) the regression models remained as
determined when all the sites were used but the interpolation was un-
dertaken leaving the selected site out. The regression jackknife gives an
indicator of the sensitivity of the regressions to outliers and the second
to extrapolation only. In the context of the data used here the biggest
issue was where the removal of a single site significantly reduced the
range of a key variable in the regression e.g. for the “North”, removal
of Darwin results in a very small latitudinal range and removal of
Malanda (Fig. 6; Cook et al., 2001) significantly reduces the elevation
range; in both of these cases the resulting regression fits a smaller pa-
rameter range increasing the geographical areas requiring extrapolation
of the regression. For this reason we believe that the regression jack-
knife is not the best indicator of the predictive ability of the isoscape,
however it does highlight the areas where additional data is required
to improve the regressions underpinning the isoscape.
The resulting annual isoscapes for δ18O, δ2H and dp are shown in Fig.
6. Large standard deviations in either jackknife approach (Supplemen-
tary Fig. S3) indicate areaswhere the predictive strength of the isoscape
areweakest, however, it is important to note that these are not errors or
uncertainties, as these cannot be determined for locations where mea-
sured data is not available. The interpolation jackknife shows that the
areas where the predictive strength of the isoscape are weakest are
those where the most depleted isotopic composition is predicted in
Fig. 6b and c along the Great Dividing Range in Central and North
Queensland, and from the Gibson Desert west of Alice Springs through
to the Gulf of Carpentaria. Whilst they are all subject to extreme events
and monsoonal influences which render the predictions of the isoscape
feasible, these are the areaswhich should be considered the highest pri-
ority for additional rainfall sampling.
There are significant spatial differences between this isoscape and
RCWIP (Terzer et al., 2013) and OIPC (Bowen and Revenaugh, 2003;
Bowen, 2018). Neither of the previous attempts reflect the depletedFig. 7. Comparison between δ2H and δ18O values from shallow, relatively young (b50 years old
different monthly rainfall amounts at three sites (represented by green stars). (a) Sydney rainfa
samples (Cendón et al., 2014b). Evaporation prior to recharge is limited and comparison with
rainfall and Georgina Basin groundwater samples (Lawn Hill, Qld; Spencer, 2009). The ground
Isa, indicating some evaporation prior to recharge. The plot strongly suggests that recharge
rainfall and Darling River Basin (Glen Villa, NSW) groundwater samples (Meredith et al., 2015precipitation observed in the Pilbara region of NW Australia (Hope
Downs) and the Gulf of Carpenteria (Mt Isa) which is largely due to
tropical cyclones and the monsoon. In particular RCWIP predicts that
the northern inland region will have extremely enriched precipitation,
presumably because of the very low annual rainfall away from the
coast, however, infrequent larger events actually dominate this zone
leading to more depleted rainfall. RCWIP also predicts precipitation in
the southern parts of Australia will bemore depleted than this isoscape,
whereas the predictions of OPIC (Bowen, 2018; Bowen and Revenaugh,
2003) are closer. Significant additional data for SE Australia in this paper
enable a higher degree of confidence in the spatial distribution for this
region in our isoscape than previously possible.
3.4. Precipitation and groundwater comparison
Groundwater studies rely on using the δ18O and δ2H of water to as-
sess its origin and recharge processes in aquifers worldwide (e.g.:
Wassenaar et al., 2009; West et al., 2014; Raidla et al., 2016). In high
rainfall areas where groundwater is recharged rapidly, it typically re-
tains the isotopic signature of rainfall. The isotopic composition of the
groundwaterwill plot close to theMWL and is close to themean precip-
itation weighted annual composition (Clark and Fritz, 1997), and may
even reflect the seasonal isotopic signature of rainfall (Jasechko et al.,
2014; Bryan et al., 2016). This recharge mechanism is seen along the
SE and SW Australian coastal fringe, e.g. the Sydney region (Cendón et
al., 2014b; Fig. 7a) where groundwater sits close to the LMWL, between
the winter and summer weighted averages. However, the groundwater
from the inland arid and semi-arid environments can reflect the epi-
sodic nature of rainfall events and/or the localised recharge of
evaporatively enriched surface water that may not be derived from
local precipitation. In these environments, the groundwater can be iso-
topically depleted relative to the annualmean rainfall, reflecting instead
the mean precipitation weighted composition of rainfall events above
some recharge threshold value (e.g. Harrington et al., 2002). In the
Georgina Basin (Spencer, 2009; Fig. 7b) groundwater is more depleted
than the weighted average rainfall and instead reflects the large rainfall
eventswhich are associatedwith cyclonic andmonsoonal activity in the
summer months, an effect seen broadly across the semi-arid northern
inland (e.g. Cendón et al., 2010; Harrington et al., 2002; Dogramaci et
al., 2012; Meredith et al., 2018). Groundwater can also be enriched
and evaporated where recharge is driven by widescale flooding) groundwater and the LMWL and precipitation amount-weighted mean compositions of
ll and Hawkesbury Sandstone (Kulnura-Mangrove Mountain Aquifer, NSW) groundwater
rainfall averages suggest that monthly rainfall N100mm leads to recharge. (b) Mount Isa
water samples plot close to or just to the right of the annual/summer LMWLs for Mount
of groundwater occurs only after intense rainfall events of N150 mm/month. (c) Cobar
). Groundwater recharge occurs during flooding originating up to 800 km away.
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 643originating hundreds or thousands of kilometres upstream (e.g.
Meredith et al., 2015; Fig. 7c). Evaporative enrichment in the unsatu-
rated zone during recharge is also well documented in arid and semi-
arid regions and has been found to strongly influence the local ground-
water isotopic composition (e.g. Barnes and Allison, 1988; Markowska
et al., 2016).
One important application of the isoscape derived in this study is to
support and guide the interpretation of δ18O and δ2H values in ground-
water studies collected from rainfall isotope data-poor regions. Compar-
ing the theoretical meteoric origin of groundwater, independent of the
recharge mechanism, with short residence time groundwater samples,
provides an opportunity to test the predictive capability of the isoscape
aswell as identifying regions of potential decoupling betweenprecipita-
tion and groundwater. Fourteen groundwater studies in northern and
eastern Australia were selected for comparison with our precipitation
isoscape (Fig. 1b; Supplementary Table S6). Most data are from Quater-
nary alluvial valleys (Baskaran et al., 2002; Cendón et al., 2010, 2014a;
DSITIA, 2014; Duvert et al., 2015; Hankin and Cendón, 2014; Herczeg,
2004; Iverach et al., 2017; King et al., 2015; Martinez et al., 2017;
Meredith et al., 2015; Please et al., 2000; Taylor et al., 2016; van der
Ley, 2016); with the exception of Lawn Hill (karst system; Spencer,
2009), Atherton Tablelands (basalt plateau; Cook et al., 2001) and
Kulnura-Mangrove Mountain (fractured sandstone Cendón et al.,
2014b). Only samples with either quantifiable tritium concentrations,
high radiocarbon (N80 pmc), and sourced from b100 m in depth have
been incorporated in an effort to ensure groundwater was representa-
tive of modern recharge (b~100 yr).
Agreement between the isoscape and the groundwater isotope sig-
nature is very close for some sites but has significant scatter around
the 1:1 line with an r2 of 0.46 for δ2H and 0.49 for δ18O. The underlying
reasons for this scatter are best investigated by comparing the isoscape
precipitation andmeasured groundwater dp (Fig. 8). Herewe see that in
high rainfall locations, groundwater dp is well predicted by the isoscape
(falling on the 1:1 line) confirming that these sites receive direct rainfall
recharge. Groundwater dp becomes progressively lower than the
isoscape dp (dashed line in Fig. 8) as aridity and distance from the
coast increases. This can be interpreted both in terms of the potential
for evaporation during recharge, but also, more importantly in Australia,Fig. 8. Comparison between dp predicted by the isoscape and that observed in modern
groundwater.the distance from the headwaters of river systems and the potential for
evaporation of surface water prior to recharge. In this dataset, a third
factor is evident for inland groundwater expected to be closer to the
dashed line, based on their geographical position:mixing of old ground-
water, discharging from underlying Great Artesian Basin (GAB) forma-
tions into alluvial valleys (Iverach et al., 2017). The GAB groundwater
component was probably recharged in distant areas closer to the coast-
line and eventuallymixeswith young locally-recharged andmore evap-
orated water. The source of water in all these cases is the same – the
Great Dividing Range, where precipitation is more depleted than seen
further inland because of the higher altitude. The difference is whether
groundwater recharges then follows a subsurface pathway to the sam-
ple location, or it is recharged locally following evaporation.
The comparison of the isoscape and groundwater stable isotope re-
cords show the predictive capability of the isoscape, particularly along
the easternmargin of the continent (Great Dividing Range) or in regions
dominated by seasonal (local) recharge processes (e.g.: monsoon). The
disconnect between the isoscape and groundwater reveals the in-
creased aridity towards the centre of the continent with sites at ~700
km(GV) and 1100 km(CC) from the coast disconnected from local rain-
fall and relying on rainfall further upstream and flooding events to pro-
duce local groundwater recharge.4. Conclusions
Australia, with an area of 7.7 million km2 ranging in latitude 12.5 to
43°S and spanning 40° in longitude, is the lowest and flattest of all con-
tinents. The original Australian GNIP network (seven stations) has been
more than doubledwith the last 8–12 years of data in the 15 station net-
work presented and interpreted in this paper.
Overall the interpretation supports the findings of Liu et al. (2010)
for coastal areas, but adds far more to our understanding of inland Aus-
tralia and many of the key water-resource limited areas such as the
Murray-Darling Basin. Local meteoric water lines were developed for
each site, as well as for the Australian continent. We recommend the
use of precipitation weighted local meteoric water lines in any hydro-
logical application of this dataset to better reflect the hydrological im-
pact of rain. When the annual precipitation weighted values were
used to generate an Australian meteoric water line, the line (δ2H =
8.3 δ18O + 14.1‰) was close to that developed from the isoscape
(δ2H= 8.6 δ18O+ 14.4‰).
A weak seasonal variation in the isotopic composition was evident
with winter minimum and spring-summer maxima at the coastal tem-
perate sites of Adelaide, Brisbane, Cape Grim, Melbourne and Sydney.
Whereas, Alice Springs and the new inland sites (Charleville, Cobar,
Mildura, Mt. Isa, Wagga Wagga, and Woomera) exhibit a spring maxi-
mum and lower values both in winter, and in summer when high rain-
fall events associated with tropical cyclones and the passage of the ITCZ
are important. Precipitation amountwas found to be a stronger driver of
isotopic composition than temperature, particularly at the northern
sites affected by tropical cyclones and the monsoon. Latitude, elevation
and distance from the coast were found to be stronger drivers of spatial
variability than temperature or rainfall amount, and the higher density
of sites provides some opportunity to look at the effect of rainout and
continentality across eastern Australia.
Annual isoscapes of δ2H, δ18O and dp developed using regional
groupings rather than climate zones allowed us to take into account
the varying importance of continentality due to either eastward or on-
shore movement of moisture and rainfall across the continent. Apart
from doubling of spatial data points, in order to develop an isoscape
other available short term rainfall data sets were used, either directly
or as check of the predictive power of the isoscape. These additional
datasets proved important in areas in the north of the continent affected
by tropical cyclones and themonsoon, as well as ones representing high
elevation areas in Australia.
644 S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645Considering the importance of groundwater resources in the driest
inhabited continent, precipitation records were compared to short res-
idence-time groundwater data. Isotopic data from alluvial or fractured
rock aquifers in the north and eastern coastal sites tend to be closely
correlated with the local rainfall predicted by the isoscape. However,
an increasing disconnect between local precipitation and groundwater
is observed inland, highlighting the importance of flooding events to
groundwater recharge. In both cases comparisons to groundwater high-
light water stable isotope precipitation records as a tool for evaluating
recharge and evaporation in local groundwater and surface water
resources.
Data statement
GNIP data used for this paper will be made available through the
Global Network of Isotopes in Precipitation database. The data and
isoscape results have been published in an institutional data repository:
Precipitation isotope data for sites in Table 1 can be downloaded
from: http://research-data.ansto.gov.au/collection/881
Gridded precipitation isoscape data can bedownloaded from: http://
research-data.ansto.gov.au/collection/880
Additional data and updates will be available at: http://www.ansto.
gov.au/ozwin
Acknowledgements
The Australian GNIP network is supported by our long-standing
partners, the Australian Bureau of Meteorology and the International
Atomic Energy Agency. Much of the historical data was analysed by
the Commonwealth Scientific and Industrial Research Organisation.
JohnGibsonprovided themuch-appreciated impetus to expand thenet-
work in 2006. Our isoscape depended on data kindly provided by Greg
Skrzypek (University of Western Australia), Steve Lewis (James Cook
University), Niels Munksgaard (Charles Darwin University), Russell
Crosbie (CSIRO), Pauline Treble and Alan Griffiths (ANSTO). Lastly we
would like to acknowledge and thank all the ANSTO people who con-
tributed to this work: Barbora Gallagher, Robert Chisari, Kellie-Anne
Farrawell, Scott Allchin, Chris Dimovski, Charmaine Day and Jenny van
Holst in the lab; Stuart Hankin for GIS support; and Pauline Treble for
commenting on a draft of the paper. This research did not receive any
grant from funding agencies in the public, commercial, or not-for-profit
sectors.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.scitotenv.2018.07.082.
References
Adomako, D., Maloszewski, P., Stumpp, C., Osae, S., Akiti, T.T., 2010. Estimating groundwa-
ter recharge from water isotope (δ2H, δ18O) depth profiles in the Densu River basin,
Ghana. Hydrol. Sci. J. 55, 1405–1416.
Araguás-Araguás, L., Froehlich, K., Rozanski, K., 2000. Deuterium and oxygen-18 isotope
composition of precipitation and atmospheric moisture. Hydrol. Process. 14,
1341–1355.9.
Baldini, L.M., McDermott, F., Baldini, J.U.L., Fischer, M.J., Mollhoff, M., 2010. An investiga-
tion of the controls on Irish precipitation δ18O values on monthly and event time-
scales. Clim. Dyn. 35:977–993. https://doi.org/10.1007/s00382-010-0774-6.
Barnes, C.J., Allison, G.B., 1988. Tracing of water movement in the unsaturated zone using
stable isotopes of hydrogen and oxygen. J. Hydrol. 100, 143–176.
Barras, V.J.L., Simmonds, I., 2008. Synoptic controls upon δ18O in southern Tasmanian pre-
cipitation. Geophys. Res. Lett. 35, L02707. https://doi.org/10.1029/2007GL031835.
Barras, V., Simmonds, I., 2009. Observation and modelling of stable water isotopes as di-
agnostics of rainfall dynamics over southeastern Australia. J. Geophys. Res. 114,
D23308. https://doi.org/10.1029/2009JD012132.
Baskaran, S., Budd, K.L., Larsen, R.M., Bauld, J., 2002. A groundwater quality assessment of
the Lower Pioneer Catchment. Qld. Bureau of Rural Sciences, Canberra.
Bowen, G.J., 2018. The Online Isotopes in Precipitation Calculator, Version 3.1. http://
www.waterisotopes.org.Bowen, G.J., Revenaugh, J., 2003. Interpolating the isotopic composition of modern mete-
oric precipitation. Water Resour. Res. 39, 1299.
Bowen, G.J., Wilkinson, B., 2002. Spatial distribution of δ18O in meteoric precipitation. Ge-
ology 30, 315–318.
Bowen, G.J., Wassenaar, L.I., Hobson, K.A., 2005. Global application of stable hydrogen and
oxygen isotopes to wildlife forensics. Oecologia 143, 337–348.
Bryan, E., Meredith, K.T., Baker, A., Post, V.E.A., Andersen, M.S., 2016. Island groundwater
resources, impacts of abstraction and a drying climate: Rottnest Island, Western Aus-
tralia. J. Hydrol. 542, 704–718.
Callow, N., McGowan, H.,Warren, L., Speirs, J., 2014. Drivers of precipitation stable oxygen
isotope variability in an alpine setting, Snowy Mountains, Australia. J. Geophys. Res.
Atmos. 119, 3016–3031.
Cartwright, I., Morgenstern, U., 2016. Contrasting transit times of water from peatlands
and eucalypt forests in the Australian Alps determined by tritium: implications for
vulnerability and the source of water in upland catchments. Hydrol. Earth Syst. Sci.
20 (12), 4757–4773.
Cartwright, I., Cendón, D., Currell, M., Meredith, K., 2017. A review of radioactive isotopes
and other residence time tracers in understanding groundwater recharge: possibili-
ties, challenges, and limitations. J. Hydrol. 555, 797–811.
Cendón, D.I., Larsen, J.R., Jones, B.G., Nanson, G.C., Rickleman, D., Hankin, S.I., Pueyo, J.J.,
Maroulis, J., 2010. Freshwater recharge into a shallow saline groundwater system,
Cooper Creek floodplain, Queensland, Australia. J. Hydrol. 392, 150–163.
Cendón, D.I., Hankin, S., Kelly, B., Hollins, S., Martel, L., 2014a. Hydrogeochemistry and Iso-
topes in Groundwater From the Condamine Catchment Alluvial Aquifer System (SE-
QLD)- Implications for Residence Times and Recharge. ANSTO/C-1395 Revision A
(Prepared for the Cotton Research Development Corporation, 59 pp and Appendices).
Cendón, D.I., Hankin, S.I., Williams, J.P., Van der ley, M., Peterson, M., Hughes, C.E.,
Meredith, K., Graham, I.T., Hollins, S.E., Levchenko, V., Chisari, R., 2014b. Groundwater
residence time in a dissected and weathered sandstone plateau: Kulnura–Mangrove
Mountain aquifer, NSW, Australia. Aust. J. Earth Sci. 62:123–141. https://doi.org/
10.1080/08120099.2014.893628.
Clark, I.D., Fritz, P., 1997. Environmental Isotopes in Hydrogeology. Lewis Publishers, New
York, p. 328.
Cook, P.G., Herczeg, A.L., McEwan, K.L., 2001. Groundwater Recharge and Stream
Baseflow: Atherton Tablelands, Queensland. CSIRO Land and Water Technical Report
08/01, April 2001.
Craig, H., 1961. Isotopic variation in meteoric waters. Science 133, 1702–1703.
Crawford, J., Hughes, C.E., Parkes, S.D., 2013. Is the isotopic composition of event based
precipitation driven by moisture source or synoptic scale weather in the Sydney
Basin, Australia? J. Hydrol. 207, 213–226.
Crawford, J., Hughes, C.E., Lykoudis, S., 2014. Alternative least squares method for deter-
mining the meteoric water line, demonstrated using GNIP data. J. Hydrol. 519,
2331–2340.
Crawford, J., Hollins, S.E., Meredith, K.T., Hughes, C.E., 2017. Precipitation stable isotope
variability and sub-cloud evaporation processes in a semi-arid region. Hydrol. Pro-
cess. 31, 20–34.
Crosbie, R., Morrow, D., Cresswell, R., Leaney, F., Lamontagne, S., Lefournour, M., 2012.
New insights to the chemical and isotopic composition of rainfall across Australia.
Water for a Healthy Country Flagship Report Report EP125581 ISSN: 1835-095X.
76pp. https://doi.org/10.5072/83/5849a0d6c685a.
Dansgaard, W., 1964. Stable isotopes in precipitation. Tellus 16, 436–468.
Dogramaci, S., Skrzypek, G., Dodson, W., Grierson, P.F., 2012. Stable isotope and
hydrochemical evolution of groundwater in the semi-arid Hamersley Basin of sub-
tropical northwest Australia. J. Hydrol. 475, 281–293.
Draxler, R.R., Rolph, G.D., 2003. Hybrid Single-Particle Lagrangian Integrated Trajectory
(HYSPLIT). model. http://www.arl.noaa.gov/ready/hysplit4.htmlN.
DSITIA, 2014. Western Cape York Groundwater Study 2. Groundwater Dependent Ecosys-
tems Investigation Supporting the Assessment of Groundwater Sustainability in the
Great Artesian Basin of Cape York. Dept Science, Information Technology, Innovation
and the Arts, Brisbane (167 pp).
Duvert, C., Raiber, M., Owen, D.D.R., Cendón, D.I., Batiot-Guilhe, C., Cox, M.E., 2015.
Hydrochemical processes in a shallow coal seam gas aquifer and its overlying
stream–alluvial system: implications for recharge and inter-aquifer connectivity.
Appl. Geochem. 61, 146–159.
Edirisinghe, E.A.N.V., Pitawala, H.M.T.G.A., Dharmagunawardhane, H.A., Wijayawardane,
R.L., 2017. Spatial and temporal variation in the stable isotope composition (δ18O
and δ2H) of rain across the tropical island of Sri Lanka. Isot. Environ. Health Stud.
1–18.
Feng, X., Faiia, A.M., Posmentier, E.S., 2009. Seasonality of isotopes in precipitation: a
global perspective. J. Geophys. Res. 114, D08116. https://doi.org/10.1029/
2008JD11279.
Froehlich, K., Gibson, J.J., Aggarwal, P.K., 2002. Deuterium excess in precipitation and its
climatological significance. Proc. Study of Environmental Change Using Isotope Tech-
niques, IAEA, Vienna, IAEA-CSP-13/P, pp. 54–66.
Gedzelman, S., Lawrence, J., Gamache, J., Black, M., Hindman, E., Black, R., Dunion, J.,
Willoughby, H., Zhang, X., 2003. Probing hurricanes with stable isotopes of rain and
water vapor. Mon. Weather Rev. 131, 1112–1127.
Goede, A., Green, D.C., Harmon, R.S., 1982. Isotopic composition of precipitation, cave
drips and actively forming speleothems at three Tasmanian cave sites. Helictite 20,
17–28.
Goede, A., Green, D.C., Harmon, R.S., 1986. Late Pleistocene palaeotemperature record
from a Tasmanian speleothem. Aust. J. Earth Sci. 33, 333–342.
Goede, A., Veer, H.H., Ayliffe, L.K., 1990. Late Quaternary palaeotemperature records for
two Tasmanian speleothems. Aust. J. Earth Sci. 37, 267–278.
Guan, H., Simmons, C.T., Love, A.J., 2009. Orographic controls on rain water isotope distri-
bution in the Mount Lofty Ranges of South Australia. J. Hydrol. 374, 255–264.
S.E. Hollins et al. / Science of the Total Environment 645 (2018) 630–645 645Guan, H., Zhang, X., Skrzypek, G., Sun, Z., Xu, X., 2013. Deuterium excess variations of rain-
fall events in a coastal area of South Australia and its relationship with synoptic
weather systems and atmospheric moisture sources. J. Geophys. Res.-Atmos. 118:
1–16. https://doi.org/10.1002/jgrd.50137.
Hankin, S., Cendón, D.I., 2014. Measurement and Preliminary Interpretation of Isotopes in
Groundwater at Talwood, QLD. ANSTO/C-1434 (34 pp).
Harrington, G.A., Herczeg, A.L., Cook, P.G., 2002. Spatial and temporal variability of ground
water recharge in Central Australia: a tracer approach. Ground Water 40, 518–528.
Harvey, F.E., 2001. Use of NADP archive samples to determine the isotope composition of
precipitation: characterizing the meteoric input function for use in ground water
studies. Ground Water 39 (3), 380–390.
Harvey, F.E., Welker, J.M., 2000. Stable isotopic composition of precipitation in the semi-
arid north-central portion of the US Great Plains. J. Hydrol. 238, 90–109.
Herczeg, A.L., 2004. Groundwater ages, sources of salt and recharge mechanisms in the
Lower Balonne area, southern Queensland, Australia: isotope and geochemical data.
CRC LEME Open File Report 164, October 2004.
Hughes, C.E., Crawford, J., 2012. A new precipitation weighted method for determining
the meteoric water line for hydrological applications demonstrated using Australian
and global GNIP data. J. Hydrol. 464–465, 42–55.
Hughes, C.E., Crawford, J., 2013. Spatial and temporal variation in precipitation isotopes in
the Sydney Basin, Australia. J. Hydrol. 489, 42–55.
IAEA/WMO, 2017. Global Network of Isotopes in Precipitation. The GNIP Database. Acces-
sible at:. http://www.iaea.org/water.
International Atomic Energy Agency, 1992. Statistical Treatment of Data on Environmen-
tal Isotopes in Precipitation: Technical Reports Series No. 331. International Atomic
Energy Agency, Vienna, p. 781.
International Atomic Energy Agency, 2014. RCWIP (Regionalized Cluster-Based Water
Isotope Prediction) Model – Gridded Precipitation δ18O|δ2H| δ18O and δ2H Isoscape
Data. International Atomic Energy Agency, Vienna, Austria 2014, available at. http://
www.iaea.org/water (downloaded at 14/1/2014).
Iverach, C.P., Cendón, D.I., Meredith, K.T., Wilcken, K.M., Hankin, S.I., Andersen, M.S., Kelly,
B.F.J., 2017. A multi-tracer approach to constraining artesian groundwater discharge
into an alluvial aquifer. Hydrol. Earth Syst. Sci. Discuss. 1–40 2017.
Jasechko, S., Birks, S.J., Gleeson, T., Wada, Y., Fawcett, P.J., Sharp, Z.D., McDonnell, J.J.,
Welker, J.M., 2014. The pronounced seasonality of global groundwater recharge.
Water Resour. Res. 50 (11), 8845–8867.
Kalbus, E., Reinstrof, F., Schirmer, M., 2006. Measuring methods for groundwater - surface
water interactions: a review. Hydrol. Earth Syst. Sci. Discuss. Eur. Geosci. Union 10
(6), 873–887.
King, A.C., Raiber, M., Cendón, D.I., Cox, M.E., Hollins, S.E., 2015. Identifying flood recharge
and inter-aquifer connectivity using multiple isotopes in subtropical Australia.
Hydrol. Earth Syst. Sci. 19, 2315–2335.
Kumar, B., Rai, S.P., Kumar, U.S., Verma, S.K., Grag, P., Kumar, S.V.V., Jaiswal, R., Purendra, B.
K., Kumar, S.R., Pande, N.G., 2010. Isotopic characteristics of Indian precipitation.
Water Resour. Res. 46, W12548. https://doi.org/10.1029/2009WR008532.
Lawrence, J.R., Gedzelman, S.D., 1996. Low stable isotope ratios of tropical cyclone rains.
Geophys. Res. Lett. 23, 527–530.
Liotta, M., Favara, R., Valenza, M., 2006. Isotopic composition of the precipitations in the
central Mediterranean: origin marks and orographic precipitation effects.
J. Geophys. Res.-Atmos. 111 (D19). https://doi.org/10.1029/2005jd006818.
Liu, J., Fu, G., Song, X., Charles, S.P., Zang, Y., Han, D., Wang, S., 2010. Stable isotopic com-
positions in Australian precipitation. J. Geophys. Res. 115, D23307. https://doi.org/
10.1029/2010JD014403.
Lykoudis, S.P., Argiriou, A.A., 2007. Gridded data set of the stable isotopic composition of
precipitation over the eastern and central Mediterranean. J. Geophys. Res. Atmos.
112, D18107.
Markowska, M., Baker, A., Andersen, M.S., Jex, C.N., Cuthbert, M.O., Rau, G.C., Graham, P.
W., Rutlidge, H., Mariethoz, G., Marjo, C.E., Treble, P.C., Edwards, N., 2016. Semi-arid
zone caves: evaporation and hydrological controls on δ18O drip water composition
and implications for speleothem paleoclimate reconstructions. Quat. Sci. Rev. 131,
285–301.
Martinez, J.L., Raiber, M., Cendón, D.I., 2017. Using 3D geological modelling and geochem-
ical mixing models to characterise alluvial aquifer recharge sources in the upper
Condamine River catchment, Queensland, Australia. Sci. Total Environ. 574, 1–18.
Matsui, E., Salati, E., Ribeiro, M.N.G., Reis, C.D., Tancredi, A.C.S.N.F., Gat, J.R., 1983. Precipi-
tation in the Central Amazon basin: the isotopic composition of rain and atmospheric
moisture at Belém and Manaus. Acta Amazon. 13 (2), 307–369.
McDermott, F., 2004. Paleo-climate reconstruction from stable isotope variations in
speleothems: a review. Quat. Sci. Rev. 23, 901–918.
Meredith, K.T., Hollins, S.E., Hughes, C.E., Cendón, D.I., Chisari, R., Griffiths, A., Crawford, J.,
2015. Evaporation and concentration gradients created by episodic river recharge in a
dryland aquifer: insights from Cl, δ18O, δ2H and 3H. J. Hydrol. 529:1070–1078. https://
doi.org/10.1016/j.jhydrol.2015.09.025.
Meredith, K.T., Han, L.F., Cendón, D.I., Crawford, J., Hankin, S., Peterson, M., Hollins, S.E.,
2018. Evolution of dissolved inorganic carbon in groundwater recharged by cyclones
and groundwater age estimations using the 14C statistical approach. Geochim.
Cosmochim. Acta 220 (Supplement C), 483–498.Moran, P.A.P., 1948. The interpretation of statistical maps. J. R. Stat. Soc. Ser. B (Methodol.)
10 (2), 243–251.
Munksgaard, N.C., Zwart, C., 2017. Pers. comm. 4/4/2017 Data From IAEA Coordinated Re-
search Project (CRP) F31004 “Stable Isotopes in Precipitation and Paleoclimatic Ar-
chives in Tropical Areas to Improve Regional Hydrological and Climatic Impact
Models.
Munksgaard, N.C., Zwart, C., Kurita, N., Bass, A., Nott, J., Bird, M.I., 2015. Stable isotope
anatomy of tropical cyclone Ita, North-Eastern Australia, April 2014. PLoS One
https://doi.org/10.1371/journal.pone.0119728.
Pfahl, S., Sodemann, H., 2014. What controls deuterium excess in global precipitation?
Clim. Past 10, 771–781.
Please, F.M., Watkins, K.L., Cresswell, R.G., Bauld, J., 2000. A Groundwater Quality Assess-
ment of the Alluvial Aquifers in the Border Rivers Catchment (Qld/NSW). Bureau of
Rural Sciences, Canberra, Australia.
Raidla, V., Kern, Z., Pärn, J., Babre, A., Erg, K., Ivask, J., Kalvāns, A., Kohán, B., Lelgus, M.,
Martma, T., Mokrik, R., 2016. A δ18O isoscape for the shallow groundwater in the Bal-
tic Artesian Basin. J. Hydrol. 542, 254–267.
Risbey, J.S., Pook, M.J., McIntosh, P.C., 2009. On the remote drivers of rainfall variability in
Australia. Mon. Weather Rev. 137, 3233–3253.
Risi, C., Bony, S., Vimeux, F., 2008. Influence of convective processes on the isotopic com-
position (δ18O and δD) of precipitation and water vapour in the tropics: 2. Physical
interpretation of the amount effect. J. Geophys. Res. 113, D19306. https://doi.org/
10.1029/2008JD009943.
Rozanski, K., Araguás-Araguás, L., Gonfiantini, R., 1993. Isotopic patterns in modern pre-
cipitation. In: Swart, P.K., et al. (Eds.), Climate Change in Continental Isotopic Records.
Geophys. Monogr. Ser. vol. 78. AGU, Washington, DC, pp. 1–36.
Scholl, M.A., Shanley, J.B., Zegarra, J.P., Coplen, T.B., 2009. The stable isotope amount effect:
new insights from NEXRAD echo tops, Luquillo Mountains, Puerto Rico. Water
Resour. Res. 45, W12407. https://doi.org/10.1029/2008WR007515.
Skrzypek, G., Mydlowski, A., Dogramaci, S., Hedley, P., Gibson, J., Grierson, P.F., 2015. Esti-
mation of evaporative loss based on the stable isotope composition of water using
hydrocalculator. J. Hydrol. 523, 781–789.
Skrzypek, G., Dogramaci, S., Rouillard, A., Grierson, P.F., 2018. Unique isotope signature of
large cyclonic events as a tracer of ecohydrologic processes in the arid subtropics, NW
Australia. J. Hydrol. (in preparation).
Spencer, J.J., 2009. The Surface and Ground Waters of the Lawn Hill Region, Northwest
Queensland. UNSW (Unpublished Honours Thesis, 97 pp).
Stern, H., de Hoedt, G., Ernst, J., 2000. Objective classification of Australian climates. Aust.
Meteorol. Mag. 49, 87–96.
Sturman, A., Tapper, N., 2006. The Weather and Climate of Australia and New Zealand.
Oxford University Press.
Tadros, C.V., Treble, P.C., Baker, A., Fairchild, I., Hankin, S., Roach, R., Markowska, M.,
McDonald, J., 2016. ENSO – cave dripwater hydrochemical relationship: a 7-year
dataset from SE Australia. Hydrol. Earth Syst. Sci. 20, 4625–4640.
Taylor, A.R., Smith, S.D., Lamontagne, S., Suckow, A., 2016. Characterisation of Shallow Al-
luvial Groundwater Resources Near Hughenden, Flinders Catchment, Queensland. A
Report to the Queensland Government Dept of Agriculture and Fisheries and the Flin-
ders Shire Council. CSIRO Report EP163793 (46 pp).
Terzer, S., Wassenaar, L.I., Araguás-Araguás, L.J., Aggarwal, P.K., 2013. Global isoscapes for
δ18O and δ2H in precipitation: improved prediction using regionalized climatic re-
gression models. Hydrol. Earth Syst. Sci. 17, 4713–4728.
Treble, P.C., Budd, W.F., Hope, P.K., Rustomji, P.K., 2005a. Synoptic-scale climate patterns
associated with rainfall δ18O in southern Australia. J. Hydrol. 302, 270–282.
Treble, P.C., Chappell, J., Gagan, M.K., McKeegan, K.D., Harrison, T.M., 2005b. In situ mea-
surement of seasonal δ18O variations and analysis of isotopic trends in a modern
speleothem from southwest Australia. Earth Planet. Sci. Lett. 233 (1), 17–32.
Treble, P.C., Bradley, C., Wood, A., Baker, A., Jex, C.N., Fairchild, I.J., Gaga, M.K., Cowley, J.,
Azcurra, C., 2013. An isotopic study of flow paths and storage in Quaternary
calcarenite, SW Australia: implications for speleothem paleoclimate records. Quat.
Sci. Rev. 64, 90–103.
Vachon, R.W., Welker, J.M., White, W.C., Vaughn, B.H., 2010. Moisture source tempera-
tures and precipitation δ18O-temperature relationships across the United States.
Water Resour. Res. 46, W07523. https://doi.org/10.1029/2009WR008558.
van der Ley, M., 2016. Hydrogeochemistry, Residence Times, and Flow Processes of
Ground and Surface Waters in the Lawn Hill Region, Northwest Queensland, Austra-
lia. UNSW (Unpublished PhD Thesis, 368 pp).
Wassenaar, L.I., Van Wilgenburg, S.L., Larson, K., Hobson, K.A., 2009. A groundwater
isoscape (δD, δ18O) for Mexico. J. Geochem. Explor. 102 (3), 123–136.
West, A.G., February, E.C., Bowen, G.J., 2014. Spatial analysis of hydrogen and oxygen sta-
ble isotopes (“isoscapes”) in ground water and tap water across South Africa.
J. Geochem. Explor. 145, 213–222.
Wu, C.F.J., 1986. Jackknife, bootstrap and other resampling methods in regression analy-
sis. Ann. Stat. 14, 1261–1295.
Zwart, C., Munksgaard, N.C., Kurita, N., Bird, M.I., 2016. Stable isotopic signature of Austra-
lian monsoon controlled by regional convection. Quat. Sci. Rev. 151, 228–235.