Introduction
Groundwater type mixing in an alluvial aquifer can occur between recently
recharged groundwater (via infiltration from the land surface) and
groundwater discharging into the alluvium from surrounding geological
formations and artesian groundwater resources (Costelloe et al., 2012;
Schilling et al., 2017; Rawling and Newton 2016; Salameh and Tarawneh, 2017).
Insufficient spatial and temporal data resolution as well as heterogeneity
in hydrogeological properties can result in considerable uncertainty when
proportioning contributions from various sources in groundwater with mixed
origins (Anderson and Woessner, 1992; Beven, 2009; Gardner et al., 2012).
Additional uncertainties confounding source attribution include change in the
magnitude of groundwater gradients and directions over time due to ongoing
groundwater abstraction (for irrigation, stock, and domestic water supplies),
and the impact that this and flood frequency may have on the extent of
artesian discharge and groundwater mixing. These complexities make it
challenging to accurately proportion contributions from various sources to an
alluvial aquifer and to guide water allocations.
Water balance modelling of alluvial aquifers is commonly used to quantify and
proportion recharge inputs from river leakage, floodwaters, areal (diffuse
recharge), and discharge from artesian sources (Anderson and Woessner, 1992;
Middlemis et al., 2000; Zhang et al., 2002; Dawes et al., 2004; Barnett et
al., 2012; Giambastiani et al., 2012; Hocking and Kelly, 2016). Historically,
hydrochemical analyses are not often used to constrain catchment-scale water
balance modelling (Reilly and Harbaugh, 2004; Barnett et al., 2012), despite
Scanlon et al. (2002) highlighting the need
to use multiple techniques (including hydrochemical insights) to increase the
reliability of recharge and discharge estimates. Geochemical data can improve
our understanding of groundwater mixing processes because of the potential to
trace pathways of groundwater movement and water–rock interactions, whilst
also providing insights into the impacts of past groundwater extractions
(Edmunds, 2009; Martinez et al., 2017). Therefore, the integration of
geochemical evidence to constrain aquifer water balance models provides a
more rigorous approach for proportioning input sources for groundwater that
has mixed origins (Raiber et al., 2015; Currell et al., 2017).
Radioactive isotopic tracers that provide insights into groundwater residence
times can constrain mechanisms of recharge and discharge, and detect
groundwater mixing. Isotopes of dissolved species can be useful for
elucidating groundwater mixing provided the different sources of groundwater
have distinctly different and consistent isotopic signatures. However, each
tracer has a different half-life and both physical and chemical processes and
calculation assumptions can affect the interpretation of groundwater
residence times (Jasechko, 2016). Therefore, multiple tracers are useful for
covering the relevant timescales and uncertainties associated with the large
range of groundwater residence times. Tritium (3H) is an excellent
indicator of modern recharge inputs in shallow groundwater (Robertson and Cherry, 1989;
Chen et al., 2006; Duvert et al., 2016), and provides valuable
information on processes active in the past ∼ 70 years.
Carbon-14 (14C) is used to understand processes active from modern to
∼ 30 ka (Clark and Fritz, 1997; Cartwright et al., 2010;
Cendón et al., 2014) and chlorine-36 (36Cl), whilst applicable in
modern groundwater (Tosaki et al., 2007), is usually reserved for the
identification of much older groundwater (100 ka to 1 Ma). One of the
challenges of using 36Cl is that, in certain cases, nucleogenic
production of 36Cl can be significant and/or varying Cl concentrations
can complicate groundwater residence time interpretations. Additionally, the
interpretation of 36Cl can be affected by the input function, as
36Cl values from rainfall vary temporally. This means that the input
function for rainfall from any time in the past may be different from current
conditions (Phillips, 2000). However, in regions with low and fairly
consistent Cl concentrations (such as in our study area), 36Cl values
can provide solid indications of old groundwater residence times (Mahara et
al., 2007).
These isotopes can also be used for tracer mixing calculations independent of
residence time estimations (Bentley et al., 1986; Andrews and Fontes, 1993;
Love et al., 2000; Moya et al., 2016). Therefore, the combination of 3H,
14C, and 36Cl dating techniques can provide hydrochemical process
insights that cannot be captured by using only one isotope.
Identification of recharge and discharge pathways (particularly from
underlying artesian contributions), and proportioning their relative
contributions in a groundwater sample can be better constrained by combining
traditional geochemical data with multiple dating techniques and other
hydrologic analyses (Amiri et al., 2016; Rawling and Newton, 2016; Schilling
et al., 2017). This is because groundwater geochemical data give insights
into long-term patterns of mixing and groundwater flow, whereas other
hydrologic data (such as hydraulic head differences) provide insights into
seasonal pumping impacts, and current local and catchment-scale groundwater
flow paths.
Here, we present for the first time a multi-tracer approach to constraining
artesian discharge from the Great Artesian Basin (GAB) into the Lower Namoi
Alluvium (LNA), north-west New South Wales (NSW), Australia (Fig. 1). We
use water stable isotopes and major ion data to assess the major recharge
and discharge pathways and occurrences of groundwater mixing in the LNA. We
also use 3H, 14C, and 36Cl to show that artesian discharge
from the underlying GAB to the LNA is locally much higher than is currently
estimated from water balance models used to guide groundwater allocations in
the region (Lower Namoi Groundwater, 2008). Our results highlight the need to
consider a multi-tracer geochemical approach when assessing artesian
contributions to alluvial aquifers and constraining water balance models of
alluvial systems globally.
Map of the study area and sample locations, along with the
location of the study area in Australia. Accompanying hydrographs show the
groundwater level response in different piezometers throughout the study
area (groundwater level data sourced from BOM, 2017). The different colours
in the hydrographs represent the different monitoring bores in the nested
set. The bottom of the slotted interval for each bore is shown in the key.
The x axis in each hydrograph is the year (1970–2010) and the y axis is
depth (between 0 and 40 m below ground surface, b.g.s.). The two locations
with red text highlight areas where the hydrograph heads show clear GAB
contribution, with the deeper piezometer showing a higher head than the
shallow one. The remaining locations show no apparent GAB contribution to
the LNA based on the hydrograph data.
Study area
The Lower Namoi River catchment is located in the north-west of NSW,
Australia (Fig. 1). Groundwater resources in the LNA are the most
intensively developed in NSW (DPI Water, 2017). For this reason, there is
concern regarding groundwater exploitation and threat to the long-term
sustainability of the system (Lower Namoi Groundwater, 2008; DPI Water, 2017).
Groundwater abstraction from the LNA supports a multibillion-dollar
agricultural sector (focused around cotton growing established in the 1960s),
supplying around 50 % of water for irrigation in the region (Powell and Scott, 2011). Peak extraction of approximately 170 × 106 m3 occurred
over the 1994–1995 growing season (Smithson, 2009). Consistently declining
groundwater levels and concern regarding the long-term sustainability of
groundwater abstraction led to the implementation of a water sharing plan in
2006, which systematically reduced groundwater allocations to the irrigation
sector over a 10-year period. The present allocation is 86 × 106 m3 year-1 (Lower Namoi Groundwater, 2008).
Hydrogeological setting
The Lower Namoi River catchment lies within the Murray–Darling Basin,
overlying the Coonamble Embayment, which is in the south-east portion of the
GAB (Radke et al., 2000). The southernmost portion of the LNA is underlain by
Triassic formations, while north-west of monitoring bore 30345 the LNA is
underlain by Jurassic formations (Fig. 2). Within the region of study, the
oldest outcropping bedrock formation is the Early Triassic Digby Formation
(lithic and quartz conglomerates, sandstones, and minor finer grained
sediments; Tadros, 1993). The Digby Formation outcrops in the south-east of
the area and the Namoi River abuts the formation just south of B' in Fig. 2. The Digby
Formation is overlain by the Triassic Napperby Formation (thinly
bedded claystone, siltstones, and sandstone). This formation occurs at a depth
of 106 m, just below the base of monitoring bore 30345 (NSW Pinneena
Groundwater Database, driller logs). In outcrops to the east of the study
area, the Napperby Formation is overlain by the Late Triassic Deriah
Formation (green lithic sandstone rich in volcanic fragments and mud clasts; Tadros, 1993). The boundary between the Triassic and Jurassic lies west of
monitoring bore 30345. The Jurassic formations important to this study are
the Purlawaugh Formation (carbonaceous claystone, siltstone, sandstone and
subordinate coal), Pilliga Sandstone (medium to coarse quartzose sandstone)
and the Orallo Formation (clayey to quartzose sandstone, subordinate
siltstone and conglomerate; Tadros, 1993). The Pilliga Sandstone forms the
bedrock below monitoring bores 25325 to 25342, and in the Namoi region is the
primary aquifer of the GAB.
A geological map of the study area and two cross sections through
the study area, showing the location and depth of the samples in the
alluvium and their proximity to formations of the GAB. Contacts obtained
from gas wells Nyora-1, Culgoora-1, and Turrawan-2, coinciding with our cross
sections are added. Their locations are displayed on the map. The general
direction of groundwater flow is from SE to NW, aligning with the B–B' line
on the map. The 36Cl data interpolated using the “natural neighbours”
algorithm is shown in each cross section.
From the Late Cretaceous to the mid-Miocene, a paleovalley was carved
through the basement rocks (Kelly et al., 2014). Then from the mid-Miocene
until present, the paleovalley was filled with reworked alluvial sediments.
Groundwater abstraction in the study area is mostly from these alluvial
sediments. Fluvial and aeolian interbedded clays, silts, sands, and gravels
form the up to ∼ 140 m thick alluvial sequence of the Lower
Namoi Catchment (Williams et al., 1989). Traditionally, three main
non-formally defined aquifers/formations have been used to describe the LNA.
The semi-confined Cubbaroo Formation overlies the bedrock in the northern
paleochannel (which passes beneath monitoring bores 25325 and 30092). This
formation is up to 60 m thick. The Cubbaroo Formation is overlain by the
semi-confined Gunnedah Formation, which is up to 80 m thick, and is
conformably overlain by the unconfined Narrabri Formation, which is 10 to 40 m
thick (Williams et al., 1989). However, recent studies in the Namoi
Catchment suggest that the rigid subdivision in to the Narrabri, Gunnedah,
and Cubbaroo formations cannot easily explain the continuum in chemical
evolution observed (discussed further below) and that the valley filling
sequence is better characterised as a distributive fluvial system (Kelly et
al., 2014; Acworth et al., 2015).
Groundwater drains from the Upper Namoi into the LNA via a bedrock
constriction north of Narrabri and generally flows from east to west within
the LNA (Barrett, 2012). Hydraulic conductivity in the alluvial aquifer is
highly variable (0.008–31 m day-1) due to the presence of variable sand and
clay (Golder Associates, 2010). However, hydraulic conductivity generally
increases with depth.
Current understanding of recharge and discharge processes in the Lower
Namoi Alluvium
There have been numerous catchment water balance models and hydrochemical
investigations in the study area because of the local and national economic
importance of the LNA. However, the hydrochemistry of the groundwater in the
region has not been used in conjunction with water balance modelling prior to
this study (Merrick, 2000; CSIRO, 2007; Kelly et al., 2007).
Water balance modelling of recharge
To guide groundwater allocations from the LNA, a series of water budget
models were developed using MODFLOW (Merrick, 2000; summarised in Kelly et
al., 2007). These models were driven by climatic, rainfall, flood, and
streamflow data and calibrated to groundwater head data. There are multiple
plausible solutions for all water balance models and the solution presented
is often constrained by several factors. These constraining factors include
geological insights, the modeller's experience and biases (such as the way diffuse recharge is modelled either as a percentage of
rainfall (Merrick, 2000; CSIRO, 2007) or as a complex evapotranspiration
function (Giambastiani et al., 2012)), verification measures, and pragmatic
goals. One MODFLOW-derived water balance model proportioned the recharge for
the water budget period 1980–1994 as the following: flood and diffuse rain
recharge, 24.1 × 105 m3 year-1; stream recharge,
33.7 × 105 m3 year-1; up-gradient alluvial inflow, 3.06 × 105 m3 year-1; and
artesian (GAB) recharge, 9.5 × 105 m3 year-1. In that model, artesian
recharge was inferred to occur in the eastern portion of the model (between
Narrabri and Wee Waa), which overlaps with this study area (Fig. 1). The
zone between Narrabri and Wee Waa accounted for 42.7 × 105 m3 year-1
of the total recharge to the LNA. Thus, according to the model, GAB discharge
into the LNA in this area was equal to 22 %. When the LNA MODFLOW model was
calibrated there was no consideration given to using hydrochemical data to
constrain the calibration (Merrick, 2000; CSIRO, 2007; Kelly et al., 2007).
Hydrochemical estimates of recharge
The first isotopic investigation in the area was conducted from 1968 to 1975
and partially published by Calf (1978). The author used 14C and 3H
to assess recharge pathways to the LNA and found evidence for river recharge
in the upper aquifer, and that modern groundwater penetrated the deeper
parts of the LNA. Calf (1978) also found evidence for “leakage” of
groundwater from the GAB up into the deeper LNA; however, volumetric
estimates were not provided.
McLean (2003) conducted an extensive hydrochemical and isotopic
characterisation of both the GAB groundwater and the alluvial groundwater in
1999–2000. This research concluded that mixing of groundwater from the GAB
into the lower and middle parts of the LNA is an important process
especially in the south of the catchment. This study also did not quantify
the amount of mixing occurring between the two groundwater sources.
The over-reliance of water balance models used to allocate groundwater
resources that have not been constrained by isotopic tracer residence times
or hydrochemical results is a common issue globally. This research
highlights that hydrochemical investigations improve our conceptual
understanding of recharge pathways and that such investigations should be
applied to all important groundwater resource assessments to enable
sustainable management.
Materials and methods
Groundwater collection
This study comprised two field campaigns, the first one from 28 January
to 8 February 2016 (summer) when the aquifer was stressed by pumping for
irrigation, and the second from 21 June to 30 June 2016 (winter) in the
absence of abstraction for irrigation.
In summer, 28 groundwater samples were collected from NSW Department of
Primary Industries Water (DPI Water) monitoring bores and a surface water
sample from the Namoi River. In winter, 16 groundwater samples were collected
from NSW DPI Water monitoring bores and surface water samples from the Namoi
River and two upstream tributaries (see Supplement Table S2 for locations).
The bores were screened at varying intervals (average length of screened
interval: 5.6 m, see Table S2 for individual bores),
intersecting the shallow, middle, and deep alluvium. Most bores were sampled
with either a Grundfos (MP1 sampling pump) or Bennett compressed air piston
pump, with the pump placed ∼ 1 m above the screen when using
the Grundfos pump. Drop-tube extensions were used with the Bennett pump to
place the pump intake just above the screen. Some deep monitoring bores were
sampled with a portable bladder pump using low-flow methods (Puls and
Barcelona, 1996). In these bores the pump was placed approximately 10 m below
standing water level, with a drop tube cut to place the pump intake within
the screen. For shallower bores (less than 50 m), a 12 V battery operated
pump was used with the pump intake placed ∼ 1 m above the
screen. For all sample sites, physico-chemical parameters (pH, DO, EC) were
monitored and samples collected once three well volumes had been pumped
and/or the physico-chemical parameters stabilised. This was generally
achieved within 1 to 3 h after onset of pumping. Sample collection
involved an in-line, 0.45 µm, high-volume filter connected to a
high-density polyethylene (HDPE) tube. Total alkalinity concentrations (field
alkalinity) were determined in the field by acid titration using a HACH
digital titrator and external pH meter control. The Fe2+ and HS-
concentrations were determined using a portable colorimeter (HACH DR/890).
Samples for anion and water stable isotope (δ2H and δ18O) analyses were collected in 60 and 30 mL HDPE bottles
respectively with no further treatment. Samples for cation analysis were
collected in 60 mL HDPE bottles and acidified with ultrapure nitric acid.
Samples for 14C and 3H were collected in 1 L narrow-mouth HDPE
bottles and 2 L HDPE bottles respectively, and were sealed with tape to
avoid potential atmospheric exchange during storage. Samples for 36Cl
were collected in 1 L narrow-mouth HDPE bottles with no further treatment.
Major ion and 14C samples were refrigerated at 4 ∘C until analysed.
We were not able to access any previously sampled GAB bores within the study
area. Thus, to better constrain GAB groundwater characteristics, we used
geochemical data from known GAB bores collected by Radke et al. (2000) and
McLean (2003). These data were collected to the north-west of our study area
and are used as a range (depending on availability of the original reported
data) for the GAB endmember in our discussions (Table S1).
To help in the description of results, we use shallow (< 30 m),
intermediate (30–80 m), and deep (> 80 m) as a rough guide to
the origin of the groundwater sample. The chosen depth categories are based
on clusters and trends in the 14C analyses.
Geochemical analyses
Groundwater samples from both campaigns were analysed at ANSTO by
inductively coupled plasma atomic emission spectroscopy (ICP-AES) for
cations and ion chromatography (IC) for anions. Samples for δ2H
and δ18O were analysed using cavity ring-down spectroscopy
(CRDS) on a Picarro L2130-i analyser. These values are reported as
per mille deviations from the international standard V-SMOW
(Vienna Standard Mean Ocean Water) and results have a precision of ±1 ‰
for δ2H and ±0.15 ‰ for δ18O.
The 14C samples were processed and analysed at ANSTO using methods
described in Cendón et al. (2014). The 14C activities were measured
by accelerator mass spectrometry (AMS) using the ANSTO 2MV tandetron
accelerator, STAR (Fink et al., 2004). The 14C results were reported as
percent modern carbon (pmc) following groundwater 14C reporting criteria
(Mook and van der Plicht, 1999; Plummer and Glynn, 2013) with an average
1σ error of 0.21 pmc.
The 3H samples were analysed at ANSTO. Water samples were distilled and
electrolytically enriched prior to analysis by liquid scintillation. The
3H concentrations were expressed in tritium units (TU) with a combined
standard uncertainty of ±0.03 TU and quantification limit of 0.04 TU.
Tritium was measured by counting beta decay in a liquid scintillation
counter (LSC). A 10 mL sample aliquot was mixed with the scintillation
cocktail that releases a photon when struck by a beta particle.
Photomultiplier tubes in the counter convert the photons to electrical
pulses that are counted over 51 cycles for 20 min.
The 36Cl / Cl and 36Cl / 37Cl ratios were measured by AMS using
the ANSTO 6MV SIRIUS Tandem Accelerator (Wilcken et al., 2017). Samples
were processed in batches of 10, with each batch containing 1 chemistry
blank. The amount of sample used was selected to yield ∼ 5 mg of Cl for
analysis without carrier addition. Chloride was recovered from
the sample solutions by precipitation of AgCl from hot solution (Stone et al., 1996).
This AgCl was re-dissolved in aqueous NH3 (20–22 wt %, IQ
grade, Seastar) to remove sulfur compounds of Ag. Owing to isobaric
interference of 36S with 36Cl in the AMS measurements, a saturated
Ba(NO3)2 solution (99.999 % trace metal basis) was used to
precipitate sulfur as BaSO4. At least 72 h were allowed for BaSO4
to settle from a cold solution (4 ∘C) in the dark before removal
of the supernatant by pipetting and filtration (0.22 Millex GS). Pure AgCl
was re-precipitated by acidifying the Ag (NH3)2-Cl solution with 5M
nitric acid (IQ Seastar, sub-boiled). Finally, AgCl was recovered, washed
twice, and dried. It was then pressed into high-purity AgBr
(99 % trace metal basis, Aldrich) in 6 mm diameter Cu target holders.
AgBr has a much lower sulfur content than Cu. The stable Cl isotopes
35Cl and 37Cl were measured with Faraday cups, and 36Cl events
were counted with a multi-anode gas ionisation chamber. Gas (Ar) stripping
(for good brightness/low-ion straggling) the ions to 5+ charge state in the
accelerator terminal suffices for effective 36S interference separation
in the ionisation chamber combined with sample-efficient and rapid analysis.
Purdue PRIMELab Z93-0005 (nominally 1.20 × 10-1236Cl / Cl) was used
for normalisation with a secondary standard (nominally 5.0 × 10-13
36Cl / Cl; Sharma et al., 1990) used for monitoring. Background
subtraction was done with a linear dependence between 36Cl rate and
interfering 36S rate. This dependency is established by combining all
the blank and test sample measurements and applied to the unknown samples
during offline data analysis. This correction factor was typically less than
analytical uncertainty of 3–4 % bar one sample that had a correction
factor of 12 % with an analytical uncertainty of 6 %.
Geochemical calculations
Calculations necessary to assess electrical neutrality, dissolved element
speciation, and saturation indices for common mineral phases were undertaken
using the PHREEQC Interactive program (3.3.8; Parkhurst and Appelo, 1999) and
the incorporated WATEQ4F thermodynamic database (Ball and Nordstrom, 1991).
The cation and anion analyses were assessed for accuracy by evaluating the
charge balance error percentage (CBE %). All samples fell within the
acceptable ±5 % range, except for samples 25327-1 (-7.8 %) and
36001-1 (-5.8 %). The inverse geochemical modelling code NEPATH XL
(Plummer et al., 1994; Parkhurst and Charlton, 2008) has been used to
calculate the mixing ratio between two endmembers, using their Cl
concentrations. The choice of endmembers will influence calculated
proportions; however, endmembers were selected to provide conservative
approximations.
Despite limitations, 36Cl residence times for selected low 36Cl / Cl
samples were calculated from the equations of Bentley et al. (1986). This
allows a direct comparison, under similar assumptions, with other estimates
obtained from GAB groundwater elsewhere (Bentley et al., 1986; Radke et al.,
2000; Love et al., 2000; Moya et al., 2016) and within the Coonamble
Embayment (Radke et al., 2000; Mahara et al., 2007). These calculations
assume a piston flow setting with no other sources or sinks besides recharge
and natural decay (Eq. 1):
t=-1λ36lnR-RseR0-Rse,
where R is 36Cl / Cl ratio measured in the sample, R0 is the
initial 36Cl / Cl ratio (meteoric water), Rse is the
36Cl / Cl ratio under secular equilibrium (in this case the 36Cl / Cl
ratio from the Pilliga Sandstone), and λ36 is the decay constant
(2.303 × 10-6). We used a R0 value of 160 (×10-15), which was
an average of 10 samples compiled from studies in the Coonamble Embayment and
reported in Radke et al. (2000). For Rse a value of 5.7 (×10-15)
was used, which is appropriate for aquifers dominated by sandstone (this
secular equilibrium value can vary according to the dominant lithology). This
Rse value has been applied to 36Cl / Cl calculations elsewhere in the
GAB (Moya et al., 2016) and is similar to that calculated from drill-core
samples recovered in the GAB by Mahara et al. (2009).
Results
Major ion chemistry
The groundwater of the alluvial aquifer is predominantly
Na-HCO3-type water,
with concentrations ranging from 0.12 to 54.6 mmol L-1 (average:
6.85 mmol L-1; SD: 8.7 mmol L-1) for Na+ and 0.29 to
24.0 mmol L-1 (average: 6.43 mmol L-1; SD: 4.8 mmol L-1)
for HCO3- (Table S2). Generally, the highest concentrations of
Na+ and HCO3- occur in the deeper groundwater and decrease up
the vertical groundwater profile (Fig. 3a). The concentration of these two
ions in the groundwater of the LNA is higher than expected from local
rainfall sources and other shallow groundwater alluvial systems in eastern
Australia (Martinez et al., 2017). In GAB groundwater, the Na : HCO3
molar ratio is generally 1:1 and the two ions are generally present in
higher concentrations than in our alluvial samples (Radke et al., 2000;
McLean, 2003), which is evident in the position of the regional GAB samples
in Fig. 3a.
Additional ions used in this study are F- and Cl- as well as the Cl / Br ratio.
The concentration of F- in the groundwater ranges from 0.002 to
0.215 mmol L-1 (average: 0.028 mmol L-1; SD: 0.04 mmol L-1). Fluoride
concentrations generally increase with depth and accumulate in solution as
all groundwater samples are below saturation with respect to fluorite (Fig. 3b).
Concentrations of Cl- in the alluvial groundwater range from 0.063 to 26.73 mmol L-1
(average: 1.67 mmol L-1; SD: 3.7 mmol L-1). Unlike
the other major ions, Cl- concentrations through the vertical
groundwater profile are relatively stable (Fig. 3c). The relationship
between Cl- and the Cl / Br ratio shows that groundwater composition
clusters from values below the seawater ratio to values close to seawater.
The Cl / Br ratios are similar to ranges found in other alluvial groundwater
systems but slightly lower than ratios observed in other GAB samples for
Australian locations (Herczeg et al., 1991; Cendón et al., 2010;
Cartwright et al., 2010). Additionally, the Cl / Br ratios in shallow samples
connected to the river are consistent with expected ratios in rainfall (Short
et al., 2017). The regional GAB samples (Radke et al., 2000) show a Cl / Br
ratio more similar to seawater, with our samples from the LNA lying on a
mixing trend between the two endmembers (Fig. 3c).
(a) Na+ vs. HCO3- showing the mixing trend that the
alluvial samples form between the Namoi River and samples from the GAB (Radke
et al., 2000; McLean, 2003). The orange calcite saturation line indicates
samples that are more enriched due to separate evapotranspiration and calcite
precipitation. The shaded blue ellipse represents all river chemistry data
available for the Namoi River and tributaries (this work, n= 4; McLean,
2003, n= 4; Mawhinney, 2011, n= 79). (b) Na+ vs. F-
and (c) Cl- / Br- vs. Cl-, highlighting the mixing trend between the
surface recharge and the GAB that we observe in other geochemical indicators.
The red dotted line represents the Cl- / Br- ratio for rainfall and
the blue dotted line is the seawater ratio.
We identified one major outlier in the hydrochemical results, which was
sample 273314. This sample is from 207 m b.g.s. and the bore screen is
classified as being in the GAB. However, the geochemical parameters for this
deep GAB sample have a signature more similar to river water than what would
be expected in the GAB 207 m b.g.s. The concentration of Na+,
HCO3-, and Cl-, F- as well as the Cl / Br ratio in this
sample plot closer to the river and shallow groundwater than the deeper
groundwater system (Fig. 3). Potential reasons for this are explored in
detail below.
Stable water isotopes (δ2H and δ18O)
The stable water isotopic values for this study range from
-0.76 to 8.4 ‰ for δ18O and
-7.5 to -54.9 ‰ for δ2H.
Most groundwater samples cluster together at around -6
and -40 ‰ (δ18O and δ2H) and lie on
the global meteoric water line (GMWL), to the right of the nearest available
local meteoric water lines (LMWL; Macquarie Marshes and Gunnedah; Fig. 4;
Table S3). A group of mostly shallow samples collected from
piezometers close to river channels define a trend to the right of the GMWL
with a slope of 5.96, which is consistent with evaporation (Cendón et al.,
2014). Our results are similar, including the shallow groundwater evaporative
trend, to those recorded by McLean (2003). Water stable isotopic compositions
for regional GAB samples range from -6.58 to
-6.24 ‰ for δ18O and -43.1 to -38.8 ‰ for δ2H (McLean, 2003; Fig. 4).
Water stable isotopes in the LNA, showing the two separate
mechanisms of recharge; surface water recharge plotting along an evaporation
trend line and potential inflow from the GAB clustered with regional samples
from the GAB (McLean, 2003).
Isotopic tracers (3H, 14C, and 36Cl)
Tritium activities vary throughout the study area, ranging from below the
quantification limit (< 0.04 TU) to 2.36 TU (average: 0.42 TU).
Tritium activities generally decrease with depth and distance from the river
channel (Fig. 5; all data in Table S3). The highest 3H
activities of 2.31 and 2.36 TU are from a sample 40 m from the river and
the Namoi River itself respectively. These are very similar to modern
rainfall in Australia (∼ 2–3 TU, Tadros et al., 2014), which
suggests modern recharge near the river channels. However, 3H
> 0.04 TU was measured at depth (down to 207 m b.g.s.). The 3H
activities we measured at depth are significant for Australian groundwater,
as the peak of the bomb pulse in Australia was only around 60 TU compared to
locations in the Northern Hemisphere. This is primarily because most
thermonuclear testing was undertaken in the Northern Hemisphere far from
Australia and mixing is limited between the atmospheric convection cells in
the Northern and Southern hemispheres. Therefore, 3H in Australian
rainfall has been at natural background concentrations for some time (Tadros
et al., 2014).
The 14C content in the groundwater ranged from 0.2 to 107.6 pmc
(average: 54.0 pmc). Generally, groundwater samples shallower than 30 m had
a high 14C content (> 90 pmc), which decreased with depth.
There were nine samples with a 14C content below 1 pmc, indicating old
groundwater (> 30 ka), with total depths ranging from 35 to
207 m b.g.s.
Our 36Cl results for the alluvial groundwater ranged from 24.06
(×10-15) to 455.35 (×10-15) (average: 169.4 (×10-15) (shown in
the interpolation in Fig. 2). It has been found that groundwater in the GAB
recharge zone closest to the study area has a 36Cl / Cl ratio up to
∼ 200 (×10-15) (Radke et al., 2000) with recharge values
applied in calculations elsewhere in the GAB of 110 (×10-15) (Moya et
al., 2016). Water from the Namoi River has a 36Cl / Cl ratio of
∼ 420 (×10-15) (Table S4).
Discussion
Identification of recharge and mixing between the GAB and the
LNA
The δ18O and δ2H isotopic compositions suggest two
mechanisms of recharge to the alluvium: artesian discharge and surface water
infiltration. The regional GAB samples plot within the alluvial groundwater
sample range, suggesting a GAB component in the alluvium. The evaporation
line in Fig. 4 indicates recharge to the alluvium via surface water
infiltration. It also shows a good connection between surface water that has
undergone evaporation prior to recharge.
Additional evidence for these two mechanisms of recharge is the composition
of Na+ and HCO3- in the LNA. Figure 3a shows a mixing line that
the alluvial samples follow, plotting between the endmembers of the GAB and
the Namoi River, suggesting an increasing GAB contribution to the alluvial
groundwater with depth. This also implies that a continuum of mixing exists
between the shallow and deep groundwater within the LNA. The shallow samples
(25220-1 and 30259-1) that are more Na+ enriched compared to samples
from the GAB have undergone separate evapotranspiration processes and hence
have a concurrent increase in Cl-. Assuming that Cl- is behaving
conservatively (Appelo and Postma, 2005) we surmise that increases in
dissolved major ion concentrations concomitant with increases in Cl- in
the shallow groundwater are likely to be a result of evapoconcentration.
Further hydrochemical evidence for these recharge mechanisms is the
covariation of Na+ and F-, both interpreted as primarily derived
from groundwater interaction with silicate minerals in this region (Airey et
al., 1978; Herczeg et al., 1991; McLean, 2003; Fig. 3b). Our alluvial samples
fall on the mixing line between samples from the river and nearby tributaries
and regional samples from the GAB (Radke et al., 2000), in a similar way to
the Na-HCO3 trend that we observe in Fig. 3a. The Cl / Br ratios in
the groundwater also support the mixing interpretation provided by the
Na+ and HCO3- concentrations, contrary to the possibility of
water–rock interactions along the alluvium flow path (Fig. 3c). Furthermore,
the relationship between 36Cl and Na+ provides additional evidence
of mixing in the groundwater (Supplement Fig. S1).
Figure 3 also highlights the deep outlying sample (273314), which was
207 m b.g.s. in total depth, yet plots with the shallow alluvial and river
samples. Figure 2 shows that this sample is situated just above the Napperby
Formation. We hypothesise that this sample originated from surface recharge
from the Namoi River (which is in contact with the underlying Digby Formation
to the south of the study area), with negligible input from the more
Na-HCO3-rich groundwater in the Pilliga Sandstone, where the sample is
from. Sample 30345-2 (Tables S2 and S3), which is situated in the lower part
of the LNA in proximity to the alluvial contact with the Napperby Formation
(Fig. 2), has similar geochemistry. These results suggest the connection
between deeper Triassic formations beneath the GAB and the Namoi River, which
must be an important consideration in future water balance models of the
catchment.
Mixing between groundwaters of varying residence times
Major ion and water stable isotope data suggest two primary mechanisms of
recharge to the LNA and show that mixing is occurring within the alluvium.
3H activity and 14C content in the alluvial groundwater quantify
the potential residence times of the groundwater sources that are mixing
within the alluvium. Tritium activities > 0.04 TU at depth (down
to 207 m b.g.s.) indicate the extent of recharge from episodic flooding.
Measuring 3H > 0.04 TU at these depths also shows that
surface recharge reaches the deeper LNA relatively quickly (< 70 years).
Tritium data from the 1970s collected from bores that were included
in our sampling campaign (25329 and 25332; Calf, 1978) suggest that 3H
was already present in the deeper parts of the alluvial aquifer (> 70 m b.g.s.)
prior to a major flood in 1971, with activities ranging from 7.9
to 11.2 TU. This indicates good connectivity to and recharge from the
surface. Additionally, measurements of 3H in these bores post-flooding
(16.6 to 20.7 TU) indicate that substantial recharge from the surface took
place during this flood.
Plot of depth vs. 3H, highlighting the 3H activity
throughout the vertical groundwater profile. Samples that fall within the
pink zone on the left are below the quantification limit
(< 0.04 TU). These data are not included in our interpretation of
how 3H changes with depth. They are presented to convey the relative
proportions of interpretable versus non-interpretable data.
This highlights the importance of surface water
recharge to the LNA. The activities of 3H > 0.04 TU
throughout the vertical profile of the LNA (Fig. 5) are inconsistent with
the low 14C contents in the groundwater. The presence of measurable
3H but negligible 14C (close to 0 pmc) suggests that mixing is
occurring between groundwater that is associated with modern recharge
processes in the alluvium and groundwater that, as indicated by the 14C
content, is presumably much older. This older groundwater may be derived from
artesian inflow. Figure 6 shows 3H activities > 0.04 TU in
samples with 14C content of almost 0 pmc, suggesting that groundwater
with very low 14C content is mixing with groundwater with high
3H activity. Even though there is evidence of 14C dilution in
localised areas, we also observe mixing between groundwaters of widely
different 14C and 3H values in the gradient of the samples in
Fig. 6 (emphasised with a dotted blue line). This gradient would be steeper
if there were mixing between groundwaters closer in residence times
(Cartwright et al., 2013).
3H (TU) vs. 14C (pmc). This shows the mixing between
groundwater with quantifiable 3H activity (as indicated by the red
band) and groundwater with very low 14C content (as indicated by the
dotted blue line).
Extent of interaction between the GAB and the LNA
The 3H and 14C values show that there is mixing between
groundwater of varying residence times; however, they provide little
constraint on the groundwaters with a 14C content of close to 0 pmc
(i.e. > 30 ka). This is where 36Cl dating can be a useful tracer
because it can be used to identify the presence of groundwaters that are
much older than the range provided by 14C.
A plot of 36Cl / Cl vs. 14C (in pmc; Fig. 7) shows a distinct
mixing trend between groundwater with high and very low 14C content. The
two deep outlying samples (30345-2 and 273314; shaded yellow ellipse in
Fig. 7) display different geochemical characteristics from the other samples,
possibly because of their proximity to the Napperby Formation (Fig. 2).
Figure 7 shows the 36Cl / Cl value range of GAB recharge,
highlighting the alluvial samples with values lower than this GAB recharge
value. Calculations suggest that these particular groundwater samples are
potentially hundreds of thousands of years old, which is consistent with
groundwater from the GAB. This implies that these alluvial groundwaters are
influenced by artesian inflow of very old groundwater. This is evident in the
natural neighbour interpolation in Fig. 2.
The apparent degree of 36Cl decay observed in the alluvial groundwater
samples is too large to be explained simply by radioactive decay as
indicated by the measurable 14C content in the same samples (Phillips,
2000). This means that the time needed for the 36Cl to decay as much as
observed would be well outside the range of 14C dating (> 30 ka)
and therefore all groundwater samples would be expected to have a
14C content of 0 pmc, which is not observed. Furthermore, the decrease
in 36Cl is unlikely to result from dilution by 36Cl-depleted
sources such as evaporites, as the Cl- concentrations are similar in
most samples (Fig. 8a and b). Therefore, mixing between groundwaters of
different residence times is the most likely explanation for the observed
36Cl signatures.
36Cl / Cl (×10-15) vs. 14C (pmc). The colour
gradient represents the mixing between the two major sources: surface water
recharge (blue is modern) and the GAB (brown is old). The shaded yellow
ellipse encompasses the two outliers where the geochemistry is being
influenced by proximity to the Napperby Formation. The shaded pink ellipse
is sample 25327-3, located in the irrigation area.
(a) 36Cl vs. Cl- concentration. The 36Cl production
arrow represents potential in situ 36Cl production from the high U and
Th content in the host rocks. (b) 36Cl / Cl ratio (×10-15)
vs.
Cl- / Br-. The dotted blue line represents the Cl- / Br-
ratio in seawater and the dotted red line represents the expected
Cl- / Br- ratio for rainfall at Narrabri based on distance from the
coast (Short et al., 2017).
Our groundwater samples from the deep alluvium display lower 36Cl / Cl
ratios (down to 24 (×10-15)) than those measured in the GAB recharge
zone. This indicates that there is very old groundwater in the deeper LNA
(conceivably older than that of the GAB recharge zone), and that the mixing
observed in our geochemical data could be taking place between groundwater
with a residence time of less than 70 years (assumed using 3H) and
groundwater with low 36Cl activities, consistent with GAB groundwater
that is potentially hundreds of thousands of years old (Radke et al., 2000).
To quantify the extent of interaction between the two groundwater sources, we
use the concentration of the conservative chloride ion to determine an
approximate percentage of GAB to alluvial groundwater at each sample
location. In general, Cl concentrations in surface water and shallow
groundwater in the study area are low (< 30 mg L-1), while samples
recovered from the Pilliga Sandstone (GAB) have higher concentrations
(∼ 60 mg L-1). To estimate the local surface infiltration
endmember, a shallow groundwater sample with high 3H activity (sample
30170-1; 2.21 TU) was used. The average of all available GAB data was used
for GAB inputs. These endmembers are mixed in varying proportions to obtain
the Cl- concentration that we observe in all our groundwater samples
(via inverse modelling calculations). If the Cl- concentration in the
sample was lower than that in the representative local surface infiltration
sample, a 100 % LNA contribution is assumed. The representative sample
used as the local surface infiltration endmember has been subject to some
evaporation and therefore does not have the lowest Cl- concentration in
the alluvium. If the sample with the lowest Cl- concentration was used
as the surface water endmember, we would have required a higher percentage of GAB
contribution across the study area. Thus, the use of the evaporated sample as
our endmember represents a conservative approach when considering the mixing
components from both the LNA and the GAB.
The Cl mixing results provide an approximate mixing threshold with shallower
samples generally containing a higher proportion of alluvial groundwater,
which diminishes with depth. These mixing proportions show that some deeper
samples in the LNA contain up to 70 % GAB groundwater. Figure 9 presents
approximate contours for artesian discharge proportions into the LNA based on
the Cl mixing approach. The dotted lines indicate areas where there is just
one sample to inform the interpretation, whereas the solid lines connect
multiple samples that all displayed similar contributions from the GAB.
Approximate percentages of GAB contribution to the LNA, calculated
from multiple geochemical tracers and major ion data.
Artesian input can be inferred from nested piezometers at locations 30481 and
30259 (Fig. 1). At these locations, the monitoring bore slotted in the
lower portion of the LNA has a head higher than the monitoring bore slotted
in the shallow portion of the LNA, indicative of upward flow. At all other
locations artesian contributions cannot be discerned from head data.
Comparing Fig. 9 to Fig. 1, we show that groundwater geochemistry can
provide a more accurate evaluation of GAB contribution to the LNA. This is
because the geochemical data can elucidate groundwater mixing processes and
provide longer-term insights compared to the hydraulic head data. Multiple
geochemical tracers reveal that boreholes in the north and west of the study
area may be experiencing much more GAB inflow than has been inferred in
catchment water balance models (Merrick, 2000; Kelly et al., 2007; CSIRO,
2007). This is most evident at sample 25342. It is not immediately apparent
from the vertical heads in the hydrograph set at sample 25342 that there is
any GAB inflow, yet based on the geochemical tracers this location is 100 %
GAB groundwater. The water balance model described in Merrick (2000)
has GAB groundwater contributing 22 % of all inflow into the LNA between
Narrabri and Wee Waa (Fig. 1). From the geochemistry alone it is not
possible to make an estimate that can be directly compared to that artesian
discharge estimate. However, it is apparent from the mixing results shown in
Fig. 9 that a large portion of the study area has an artesian input to the
LNA that is likely to be greater than 22 %. The above observations
highlight why geochemical insights should ideally be used as one of the
constraining data sets when doing water balance models in regions where there
is both artesian discharge and surface water recharge to the alluvial
aquifer.
Changes in 14C content (pmc) in select boreholes in the study
area between 1978 and 2016 (see Figs. 1 and 8 for the locations of the
bores). The five bores in bold text highlight where we observe changes in the
14C content from 1978 to this study. Where available, the season and/or
year of sampling is included. ND: no data.
Bore
Depth interval
Calf
McLean
ANSTO data
This study
This study
(m b.g.s.)
(1978)
(2003)
(summer 2010)
(summer 2016)
(winter 2016)
25220/1
24.4–30.5
28.15
ND
ND
69.66
69.94
25220/3
97.5–109.7
0.99
ND
0.13
0.17
0.22
25325/2
36.9–38.4
83.63
ND
85.77
86.25
ND
25325/6
67.1–70.1
65.31
ND
66.57
90.37
ND
25332/1
17.7–21
103.61
ND
ND
102.48
ND
25332/2
38.1–41.1
99.19
ND
104.78
ND
ND
25332/3
50.9–55.5
94.70
ND
ND
ND
ND
25332/4
66.8–69.8
49.33
ND
84.12
73.57
ND
25327/1
18.9–21.9
123.36
101.3 (s)
ND
103.43
102.74
25327/2
57.9–60.9
84.16
93.78 (s)
ND
92.05
90.56
25327/3
80.8–83.8
8.48
8.63 (s)
ND
25.79
56.08
30092/1
17.7–20.7
ND
90.51 (w)
ND
ND
ND
30092/2
48.2–49.4
ND
80.06 (w)
72.31
ND
66.92
30092/4
108.2–110
ND
0.19 (w)
0.24
0.3
0.21
Temporal changes in the interaction between the LNA and the GAB
The multiple geochemical tracers we have used show substantial artesian
discharge to the LNA, which is larger than that currently considered in
groundwater models of the region (Merrick, 2000; Kelly et al., 2007; CSIRO,
2007). However, it is difficult to constrain how the extent of artesian
discharge has changed over time and how it may continue to change. Time
series sampling can constrain how this GAB discharge will change and is
important for understanding future artesian contributions to the LNA. Past
14C (pmc) data collected from the same bores in 1978 (Calf),
2003 (McLean), 2010 (unpublished ANSTO data), and 2016 (this study) enable us to
observe how the 14C content in the groundwater has changed over time.
The historical 14C data, coupled with data from this study, have the
potential to be used as a preliminary indicator of changes in the relative
contributions of high 14C contents from recent surface recharge
(∼ 100 pmc) versus low 14C contents of the GAB discharge to the
LNA. The data set contains 14 bores from five nested sites and is the most
comprehensive long-term time series database for the study area, if not all
of Australia, despite not being complete for all years.
Most of the samples displayed relatively consistent 14C values across
the years where data were available. However, we observed large changes in
14C content in five monitoring bores; four showed an increase and one showed a
decrease (bold text in Table 1). This suggests that the varying contributions
of older and younger groundwater has changed over time, which could be a
preliminary indicator of increased surface recharge to various sites, or
increased artesian discharge to others. Therefore, measuring the 14C in
the groundwater at any future time and assessing how this has changed using
past data is useful as a preliminary indicator for the current state of the
system. However, consistent data collection and incorporation of other
factors that may affect groundwater mixing (such as rate of groundwater
extraction and amount of surface infiltration) are necessary to make
inferences about temporal changes in the interaction between the LNA and the
GAB.
Conclusion
We have used multiple geochemical tracers to show that artesian discharge to
a shallow alluvial aquifer is higher than previously derived from water
balance models in the literature (Merrick, 2000; CSIRO, 2007; Kelly et al.,
2007). We have also provided a percentage estimate of GAB groundwater in each
sample collected in the LNA using the concentration of Cl in the groundwater,
showing that in some locations the “alluvial” sample is comprised of up to
70 % GAB groundwater. Our findings are important when considering the
global importance of groundwater and the sustainable use of connected
alluvial and artesian systems, globally.
Isotopic tracers (3H, 14C, and 36Cl) indicate that there is
substantial mixing between two groundwater endmembers of very different
residence times (< 70 years and very old groundwater consistent with
the GAB). This suggests interaction between modern surface recharge through
the shallow LNA and variable artesian inflow at depth, dependent on where the
sample is located in the system. We have also used past 14C data (1978,
2003, 2010), along with data from this study to show that these data can be
used as a preliminary indicator of how the extent of interaction between the
GAB and the LNA has changed over time. Yet, how these trends change
geographically throughout the system and how they will behave in the future
are difficult to constrain without continuous monitoring.
In the interval of the Lower Namoi studied discharge from the GAB into the
LNA was previously considered to contribute approximately less than 22 %
of the input water to the LNA (Merrick, 2000; CSIRO, 2007; Kelly et al., 2007).
However, the geochemical data reported above clearly indicate that GAB
discharge is occurring in locations where inflow is not apparent from the
nested hydrograph data. This highlights the need to apply multiple
groundwater investigation techniques (including flow modelling, hydrograph
analysis, geophysics, and geochemistry) when inferring artesian discharge to
an alluvial aquifer. This research has demonstrated that a multi-tracer
geochemical approach is required to better determine artesian contributions
to the alluvial aquifer and must be considered in constraining future models
of the study system and elsewhere.