Science of the Total Environment 574 (2017) 1–18

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Using 3D geological modelling and geochemical mixing models to

characterise alluvial aquifer recharge sources in the upper Condamine
River catchment, Queensland, Australia
Jorge L. Martinez a,⁎, Matthias Raiber b, Dioni I. Cendón c,d
a
Queensland University of Technology, Brisbane, Australia
b
CSIRO Land and Water, Brisbane, Australia
c
Australian Nuclear Science and Technology Organisation, Lucas Heights, Australia
d
Connected Water Initiative, School of Biological, Earth and Environmental Sciences, University of New South Wales (UNSW), Sydney, Australia

H 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

• Approach integrates 3D geological

modelling with EMMA and mixing cal-
culations
• Successful characterisation of alluvial
recharge sources and controlling factors
• Four end-members identified: rainfall
and groundwater inflow from adjacent
aquifers
• High contribution rates of mountain
front recharge to alluvial
hydrochemistry
• Mixing, evapotranspiration and water-
rock interaction control groundwater
evolution

a r t i c l e i n f o a b s t r a c t

Article history: The influence of mountain front recharge on the water balance of alluvial valley aquifers located in upland catch-
Received 7 June 2016 ments of the Condamine River basin in Queensland, Australia, is investigated through the development of an in-
Received in revised form 4 September 2016 tegrated hydrogeological framework. A combination of three-dimensional (3D) geological modelling, hydraulic
Accepted 4 September 2016
gradient maps, multivariate statistical analyses and hydrochemical mixing calculations is proposed for the iden-
Available online 10 September 2016
tification of hydrochemical end-members and quantification of the relative contributions of each end-member to
Editor: D. Barcelo alluvial aquifer recharge.
The recognised end-members correspond to diffuse recharge and lateral groundwater inflows from three
Keywords: hydrostratigraphic units directly connected to the alluvial aquifer. This approach allows mapping zones of poten-
EMMA tial inter-aquifer connectivity and areas of groundwater mixing between underlying units and the alluvium.
Netpath Mixing calculations using samples collected under baseflow conditions reveal that lateral contribution from a re-
Hydrogeochemistry gional volcanic aquifer system represents the majority (41%) of inflows to the alluvial aquifer. Diffuse recharge
MIX contribution (35%) and inflow from two sedimentary bedrock hydrostratigraphic units (collectively 24%) com-
prise the remainder of major recharge sources.
A detailed geochemical assessment of alluvial groundwater evolution along a selected flowpath of a representa-
tive subcatchment of the Condamine River basin confirms mixing as a key process responsible for observed spa-
tial variations in hydrochemistry. Dissolution of basalt-related minerals and dolomite, CO2 uptake, ion-exchange,

⁎ Corresponding author.
E-mail address: jorgelmarti@gmail.com (J.L. Martinez).

http://dx.doi.org/10.1016/j.scitotenv.2016.09.029
0048-9697/© 2016 Elsevier B.V. All rights reserved.
2 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18

precipitation of clay minerals, and evapotranspiration further contribute to the hydrochemical evolution of
groundwater in the upland alluvial aquifer.
This study highlights the benefits of undertaking an integrated approach that combines multiple independent
lines of evidence. The proposed methods can be applied to investigate processes associated with inter-aquifer
mixing, including groundwater contamination resulting from depressurisation of underlying geological units hy-
draulically connected to the shallower water reservoirs.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction
In semi-arid regions, a substantial proportion of valley alluvial aquifer
replenishment occurs as mountain front recharge (Voeckler, 2012). This
process is the result of a combination of various factors; e.g., occurrence
of thinner soil profiles, higher orographic/hydraulic gradients and lower
evapotranspiration potential, commonly associated with such mountain-
ous terrains (e.g.Wilson and Guan, 2004). Even under more humid condi-
tions, aquifers positioned in lower parts of catchments also receive
contributions from mountain front systems along the mountain-plain
transition zone (Liu and Yamanaka, 2012; Mikita et al., 2011).
Alluvial valley aquifers are usually hydraulically connected to the
subsurface component of mountain front recharge, commonly named
mountain block recharge (Aishlin and McNamara, 2011). The identifica-
tion of mountain block recharge sources and the understanding of spa-
tial variability in infiltration rates due to the influence of climatic
variability and land use changes to catchments are key elements for sus-
tainable management practices of water resources.
In groundwater resource investigations, three-dimensional (3D)
geological models have historically been used to provide the model
structure for numerical modelling simulations (Robins et al., 2005).
However, the increased availability of digital geological datasets and
low- and high end modelling software packages together with im-
proved hardware capabilities allow the development of 3D geological
models for a wider variety of hydrogeolological objectives, including
the development of improved hydrogeological conceptual models
(Giambastiani et al., 2012; Kelly et al., 2014; King et al., 2014; Moya
et al., 2014; Raiber et al., 2015; Raiber et al., 2012).
In many hydrogeological settings, the chemical composition of a
groundwater sample can be derived from mixing of water from differ-
ent sources with distinct chemical compositions. Inter-aquifer ex-
changes and groundwater interaction with surface water and rainfall
are commonly the major factors controlling the hydrochemical evolu-
tion of shallow groundwaters (Gómez et al., 2014), followed by evapo-
transpiration, water-rock interactions and/or microbiological activities.
Mixing between aquifers is controlled by several factors, including
the geometry and lithology of the geological host formations. Mixing
zones between aquifers can occur, for example, where deeper aquifers
‘pinch out’ or sub crop beneath shallower aquifers, or where faults im-
pede horizontal groundwater flow, e.g. Moya et al. (2014); Cartwright
et al. (2010).
Groundwater sources involved in mixing can be characterised as
end-members and their relative contribution to a mixed sample can
be estimated using mixing models (Doctor et al., 2006). End-member
mixing analysis (EMMA) uses Principal Component Analysis to identify
the minimum number of end-members necessary to explain the vari-
ability of a set of chemical species in a group of samples (Tubau et al.,
2014). This approach has been explored for identification of stream
water recharge sources (Christophersen et al., 1990; Hooper, 2003), in
karst environments (Doctor et al., 2006; Miller et al., 2015), coal mine
groundwater (Sun and Gui, 2015) and at different catchment scales
(James and Roulet, 2006).
Mixing models have been used in groundwater studies for a variety of
purposes, including the quantification of vertical intra- and inter-aquifer
exchanges (Gómez et al., 2014; Morales-Casique, 2012; Rueedi et al.,
2005) and to estimate the degree of interaction between surface and
groundwater (Kalbus et al., 2006; Oyarzún et al., 2014). The selection of
tracers for such calculations ranges from the use of conservative species
such as chloride and stable isotopes of water (Cendón et al., 2010; Liu
and Yamanaka, 2012; Nakaya et al., 2007; Vanderzalm et al., 2011) to a
set containing multiple tracers including major ions, minor elements (F,
Li), and various isotopes (e.g. δ13CDIC, δ34SSO4, and δ18OSO4) (Jurado et al.,
2013; Ramos-Leal et al., 2007). For example, multi-tracer approaches
have been used to estimate the contribution of anthropogenic recharge
sources to shallow aquifers in urban settings (Jurado et al., 2012;
Vázquez-Suñé et al., 2010).
Several authors highlighted that for successful identification of re-
charge sources to an aquifer system and the estimation of their relative
significance, such sources need to be underpinned by a solid conceptual
model of a regional hydrogeological system. The model should represent
the understanding of the hydrological dynamics and hydrochemical pro-
cesses and is ideally based on multiple independent lines of evidence
(Gómez et al., 2014; Morales-Casique, 2012; Raiber et al., 2015; Tubau
et al., 2014). Hydraulic data can greatly assist in the identification of
flow paths and zones where possible inter-aquifer exchanges occur,
whereas the interpretation of hydrochemistry can guide the identification
of end-members and selection of appropriate chemical constraints for
mixing calculations (Vázquez-Suñé et al., 2010).
The objectives of this study are to:
(i) develop a hydrogeological conceptual model of the headwaters of
the Condamine River basin from geological and hydrogeological data, in
order to identify zones of inter-aquifer connection;
(ii) use end-member mixing models to estimate relative inflow con-
tributions to the alluvial aquifer in the headwaters of the Condamine
River basin; and
(iii) identify physical, geomorphological and chemical processes af-
fecting the hydrochemical evolution of alluvial groundwater by examin-
ing spatial variations along a single representative flow path.
2. Study area background
2.1. Climate and land use
The upper reaches of the Condamine River catchment are located
approximately 200 km west of Brisbane in southeast Queensland,
Australia. The study area covers approximately 6000 km2 and comprises
five major tributary systems located east of the main Condamine River
channel (Fig. 1).
Long-term rainfall data of several gauging stations within the study
area (Bureau of Meteorology, 2015) indicate a seasonal trend with
highest rainfall months between November and March. Approximately
40% of annual precipitation, with an average of 665.5 mm (Clifton post
office station), occurs between December and February (Martinez
et al., 2015). However, the spatial distribution of rainfall is strongly in-
fluenced by an orographic effect caused by the steep escarpments of
the Great Dividing Range to the east, where ground elevation can
reach up to 1300 m above Australian Height Datum (AHD) with a grad-
ual decrease to the west reaching an average altitude of 400 m AHD in
the main drainage line of the Condamine River. In the upper reaches
of the tributaries, both annual mean rainfall and evapotranspiration
are approximately 1600 mm, whereas in the central to northwest
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18 3

Fig. 1. – Study area with identification of samples used in the mixing calculations and selected end-members.

portion, average rainfall and evapotranspiration values are around
800 mm and 2000 mm, respectively (Bureau of Meteorology, 2015).
Up to 90% of groundwater extracted from the alluvial aquifer system
is used for irrigated agriculture in the central parts of the catchment
(Tan et al., 2012). Pastures cover the majority of the catchment area
(96%) however irrigated fields occupy 75% of the alluvial footprint.
With the exception of two tributaries located in the north of the area
(i.e., the Hodgson and Kings creeks, Fig. 1), the alluvial aquifer repre-
sents the most developed groundwater resource in the basin, compris-
ing 44% of all registered bores.
2.2. Regional geology
The boundaries of the Condamine River catchment in the study area
intersect two major geological domains. To the east the Condamine
River catchment overlies the Triassic-Cretaceous Clarence-Moreton
Basin (Fig. 1). The Clarence-Moreton Basin is connected to the
Jurassic-Cretaceous Surat Basin through the Toowoomba Strait and
across the Kumbarilla Ridge, (Jell and Queensland, 2013). To the
west the Clarence-Moreton Basin laps onto the Palaeozoic bedrocks
of the New England Orogen (Texas Sub province) and mixed-
volcanic sedimentary and intrusive rocks of Palaeozoic and Mesozoic
ages.
Major geological structures in the region include a series of north-
west trending basement faults, which are up to 40 km long and are pre-
dominantly associated with the Early Carboniferous rocks of the Texas
Beds formation. The only significant structure mapped in the
Clarence-Moreton Basin within the study area is the Swan Creek anti-
cline, with axis and reverse faults of northwest orientation and fault
dips of over 45°.
4 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
3. Methods
A hydrogeological conceptual model was developed using long-term
hydraulic and hydrochemical data sets to serve as a basis for mixing cal-
culations and the identification of processes responsible for the
hydrochemical evolution of groundwaters. End-member mixing models
were tested using two datasets and mixing ratios were computed con-
sidering most likely end-members identified using both historical and
recent data. Relative contributions of the inflow sources to the alluvial
aquifer were investigated and the spatial variability of these results
was analysed within the 3D framework. Chemical reactions and physi-
cal processes responsible for the hydrochemical evolution of alluvial
aquifer were investigated via a mass balance reaction model of a repre-
sentative groundwater flow path.
3.1. 3D geological model
A regional scale 3D geological model of the Condamine River basin
was developed using GoCAD™ and SKUA (Paradigm Geophysical Pty
Ltd) 3D geological modelling software to integrate stratigraphic and
lithological data. The model development followed the workflow de-
scribed by Raiber et al. (2012) and Moya et al. (2014). Key steps includ-
ed data validation, the use of control points to assist the definition of
contact surfaces between geological units, and the generation of 3D
structures from 2D boundary surfaces.
A variety of data sources (Appendix A) was combined to estimate
the contact surfaces between different geological units. Lithological
data from 7659 bores was revised and validated against surface geolog-
ical maps, the local stratigraphic column (QDME, 2008), digital eleva-
tion models and interpreted cross-sections. The resulting dataset
contained lithological and stratigraphic descriptions of 4987 boreholes,
with depth intervals being grouped and attributed to five
hydrostratigraphic units (HSU). The definition of HSUs and criteria
used for their characterisation is described in Section 4.1.
3.2. Groundwater data interpretation
Spatial distributions of hydraulic head were produced for each of the
HSUs using time series of hydraulic head measurements sourced from
the groundwater database (GWDB), held by the Queensland Depart-
ment of Natural Resources and Mines (DNRM, 2013a).
Data for the period between 1940′s and 1995 were aggregated as
median values when the difference between the maximum and mini-
mum observed values was less than or equal to 0.5 m. Moreover, the
depth of water intake (screen interval) at each bore assigned to a specif-
ic geological formation in the GWDB was cross-checked in order to
match one of the five HSUs considered in the study. Bores without infor-
mation of water intake depth or signs to cross more than one HSU were
not included in the interpretation.
Potentiometric surfaces were interpolated using ordinary kriging
(Stein, 1999; Tonkin and Larson, 2002) and vertical hydraulic gradients
between the alluvial aquifer and underlying HSUs were estimated with-
in Leapfrog and ESRI ArcMap platforms. This approach was used to iden-
tify groundwater flow paths and zones of interaction between the
alluvial aquifer and underlying HSUs.
Although the potentiometric surfaces were derived from data sets
with a relatively high spatial density, it should be noted that effects of
groundwater extraction were not taken into account in their analysis.
Nevertheless, since the investigation aims at regional scale processes,
local effects are unlikely to affect the interpretations.
3.3. Sampling, analytical procedures and hydrochemical datasets
Historical groundwater hydrochemical data for the period
1971–2011 were obtained from the GWDB (DNRM, 2013a). Historical
surface water hydrochemical data for the period 1999–2013 were
sourced from the DNRM surface water database (SWDB) (DNRM,
2013b). At sites featuring multiple observations, median values were
calculated.
In addition, a set of 97 samples (two composite rainfall, 65 ground-
water and 30 surface water) collected mid-2014 and mid-2015 were
tested in situ for electrical conductivity (EC), pH, oxidation–reduction
potential (ORP), temperature, dissolved oxygen (DO) and analysed for
major ion and selected minor ion concentrations. Charge balance errors
were within an acceptable range (i.e. ≤ 5%). Stable isotopes of water
were reported with 0.5‰ precision for δ2H and 0.1‰ for δ18O. Sampling
and analytical procedures undertaken were described in Martinez et al.
(2015) and a complete set of analytical results is presented in Appendix
B.
3.4. Multivariate statistical analysis and end-member mixing analysis
(EMMA)
Principal Component Analysis (PCA) was used in combination with
Hierarchical Cluster Analysis (HCA) (Monjerezi et al., 2011) to develop
an understanding of the main processes controlling the groundwater
and surface water hydrochemistry in the study area. Findings from
these multivariate statistical analyses supported the design of a prelim-
inary hydrochemical conceptual model required for the application of
mixing calculations.
The HCA procedure was described in detail by Martinez et al. (2015)
and was applied to a combined set of samples derived from precipita-
tion, groundwater and surface water. Due to the high frequency of re-
peated samples from individual gauging stations, surface water results
were separated into wet and dry periods (Nov–Mar and Apr–Oct, re-
spectively) and the median values observed at each sampling site
were used.
PCA was used to reduce the total number of analytical variables and
to investigate the degree of correlation between samples with respect to
hydrostratigraphic units and hierarchical clusters. In addition, PCA was
also employed to obtain statistical factors for the EMMA, assisting in
the identification of the minimum number of end-members within
the dataset.
End member mixing analysis was initially undertaken using a
dataset comprising 928 groundwater samples, five rainfall samples, 15
surface water samples collected during ‘wet’ periods and 44 surface
water samples collected during ‘dry’ periods. Following the ‘rule of
one’ (Hooper, 2003; Tubau et al., 2014), only eigenvectors (or principal
components) that explained a percentage higher than a value obtained
via “100/n. ° of variables” were retained for the EMMA. In order to assess
the sampling adequacy and the validity of using PCA for EMMA,
Bartlett's sphericity test and the Kaiser-Meyer-Olkin (KMO) test (Field,
2009) were carried out using SPSS v.21 for Windows (IBM). A total of
12 variables were tested in different combinations (tests), including
physical-chemical parameters (EC, pH), major cations (Na, K, Ca, Mg),
major anions (HCO3, Cl, SO4, NO3), and other constituents (F, SiO2). Ex-
cept for pH, all parameters were log10 transformed to approximate nor-
mal distribution and the data were standardised prior to the analysis.
Following the findings of the preliminary EMMA using the large his-
torical dataset, EMMA workflow was reapplied to a dataset from mid-
2014 and mid-2015, which included 65 groundwater samples, 30
surface water samples and two rainfall samples. These samples were
collected under similar meteorological conditions which eliminate pos-
sible seasonal variability in end-member composition.
The number of tracers required for mixing ratio computation using a
mass balance approach is proportional to the number of potential
sources. From the minimum number of end-members identified by
EMMA (i.e., ne), the selection of tracers was based on the minimum
compliance of the number of species recommended by Vázquez-Suñé
et al. (2010), or ne-1, the deduction of their behaviour (conservative ver-
sus reactive) and the overall groundwater balances.
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
3.5. Mixing calculations
While the mixing of ni conservative solutes from ne end-members is
a linear process in nj samples, non-linear factors may be introduced by
thermodynamic activities, which are affected by water-rock interac-
tions (Rezaei et al., 2005). Solute mixing models are typically based on
the conservation of mass and assume that the mixing proportions of
all end-members sum to unity (De Gaspari, 2015).
A maximum likelihood algorithm developed by Carrera et al. (2004)
allows evaluation of the proportion in which waters from different
sources are mixed in a particular sample, further accounting for uncer-
tain end-members. The uncertainty is reduced by including the mixed
samples in the calculations and by assigning standard deviation values
to end-members and mixed samples. This approach code was revised
by Vázquez-Suñé et al. (2010), who used eigenanalysis to constrain
the mixing ratios of end-members at particular sample locations.
Sources of uncertainty associated with mixing calculations include
errors related to sampling and analysis, influence of thermodynamic
processes, spatial and temporal variability in hydrochemical end-
members composition, the lack of data, or uncertainty with regards to
the selection of end-members (Jurado et al., 2015; Morales-Casique,
2012). To minimise these uncertainties, tests using (1) different stan-
dard deviation values for both end-members and samples and
(2) fixed mixing ratios for particular samples (Vázquez-Suñé et al.,
2010) were conducted.
Following the identification of potential sources of inflow to the allu-
vial aquifer (i.e., end-members) and the selection of adequate tracers
(both conservative and reactive species), mixing ratio computations
were carried out using the MIX code (http://www.h2ogeo.upc.edu/
software/MIX PROGRAM/index.htm; Vázquez-Suñé et al. (2010)). In
order to assess the end-member ratios at each sample, the following
equation was used (Jurado et al., 2015):
Xij ¼ ∑ λej Yie
e
where Xij and Yie correspond to the concentration of solute i in sample j
and end-member e, respectively, and λej relate to the proportion of end-
member e in mixture j.
Furthermore, the following constraints must be satisfied:
0≤λej ≤1
X generally found at the base of the channels and can extend to lower
λ ¼1
e ej
The hydraulic gradient maps were then used to identify sites at
which samples could constrain mixing ratios. The evaluation of end-
members and selected tracers was conducted through a sensitivity anal-
ysis which assessed various sources of alluvial aquifer inflow. Results
were subsequently interpreted through comparisons to the
hydrogeological conceptual model.
3.6. Mass balance reaction model of alluvial aquifer
The inverse geochemical model code NETPATH XL (Parkhurst and
Charlton, 2008; Plummer et al., 1994) was employed to examine the
hydrochemical evolution along a representative groundwater flow
path within the alluvial aquifer. Such processes included chemical reac-
tions associated with water-rock interaction and the influence of evap-
oration. The flow path identified is located within the Dalrymple Creek
tributary (Fig. 1), which encompasses the major geological and geomor-
phological domains. The flow path was divided into two segments:
(FP1a) the upper recharge zones, where the alluvium overlies MRV
and is expected to highly interact with this HSU; and (FP1b) in the
lower floodplains, where evapotranspiration and mixing between the
alluvial aquifer and underlying HSUs are more likely to occur. Dividing
5
the flow path into two segments enabled mixing ratios to be better
constrained. Mass balance reactive model results were also more con-
sistent with changes in lithology along the flow path (Appendix D).
Mineral phases used in the model were based on the basalt-
dominant MRV mineralogy (Free, 1989), as this unit is the major source
of alluvial sediments in the upper segment of the tributary. Investiga-
tion bores drilled at the base of the lower segment further assisted in
the selection of minerals and constraints in NETPATH XL.
Analytical results of the samples from mid-2014 and mid-2015 were
used as initial concentrations for both segments, using end-member
contribution ratios at similar proportions of that obtained from linear
mixing models (Section 4.6).
4. Results and discussion
4.1. Hydrogeological conceptual model
Based on the extents of geological formations mapped at 1:100,000
scale (QDME, 2008) and hydraulic properties of aquifers and aquitards
previously characterised on broader scales (Huxley, 1982; Klohn
Crippen Berger, 2014; Lane, 1979; Lewis et al., 2008; Martinez et al.,
2015), five HSUs were defined. The HSUs (from youngest to oldest) cor-
respond to: Alluvium, Main Range Volcanic (MRV), Walloon Coal Mea-
sures (WCM), Marburg Subgroup (MS), and Palaeozoic basement.
At the regional scale, not all HSUs behave as homogeneous aquifers
or leaky aquitards and may comprise a sequence of smaller water bear-
ing zones. This is due to intercalation of varied grain size and different
degrees of cementation, controlling the amount of groundwater stored
and available in these units. Nevertheless, this approach does serve
the aim to assess the contribution of mountain block recharge to the al-
luvial aquifer in the scale chosen for the study.
The lithological composition, vertical extent and hydraulic proper-
ties of each HSU are summarised in Fig. 2 and the 3D distribution of
ð1Þ HSUs is depicted in Fig. 3.
The alluvial aquifer occupies approximately 16% of the study area,
and overlies the other four HSUs. The alluvium is up to 70 m thick
along the main river channel, including 50 m thickness in the Dalrymple
Creek tributary. Other tributaries have a maximum alluvial thickness of
30 m (Glengallan and Swan Creek) and 20 m (Hodgson, The Head). In
the headwaters, the alluvial tributary systems are deeply incised into
ð2Þ the underlying volcanic bedrock. Channel fill sediments are generally
coarse grained and derived from the MRV. These coarse sediments are
ð3Þ
lying areas of the catchment. In these lower sections of the tributaries,
thick black soils (average thickness of 1.5 m) overlay clay or sandy
clay (~thickness of 10 m) interbedded with thinner layers of medium
to coarse sand (2–5 m) mostly associated with abandoned channels,
which are likely to occur as non-continuous lenses. The alluvial sedi-
ments are poorly consolidated, with some localised occurrences of calci-
um carbonate and iron oxide (Huxley, 1982). The level of lithological
heterogeneity identified in the alluvial aquifer is reflected by the wide
range of hydraulic conductivity values as shown in Fig. 2. Groundwater
flow in the alluvium is generally driven by the topography and thickness
of this unit. The regional scale potentiometric surface in the alluvium
suggests that groundwater flows from the upper (i.e., eastern–south-
eastern) parts of the catchment to the lowlands in the centre and north-
west of the basin, with a mean hydraulic gradient of 0.003 at Dalrymple
Creek.
The MRV fractured rock aquifer outcrops along the eastern and
north-eastern portions of the study area (Fig. 3). It is a highly developed
aquifer in the area, providing up to 80% of groundwater extracted in the
Hodgson Creek and Kings Creek subcatchments (English et al., 2008).
The extrusive rocks of the MRV aquifer correspond to a sequence of
up to 16 individual lava flows (Willey, 2003). Weathering horizons have
developed at the paleo-surface in between individual lava flows, and
ash deposits and sedimentary deposits interbedded between basalt
6 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18

Fig. 2. – Hydrostratigraphic column of the Condamine River Basin including physical and hydraulic properties of each unit. Lithological composition description was based on Fielding
(1993), Huxley (1982), Free (1989) and interpretation of alluvial sediment drilling cores. Notes: 1 Kh – horizontal permeability coefficient; 2 within the domain of the study area; 3
within the domain of the Surat Basin; 4 data from Hutton Sandstone (equivalent to Koukandowie Formation in Marburg Subgroup).

flows are also present and influence groundwater storage and flow intensive network of vertical joints and fractures, which represent
(Free, 1989). Vesicular zones within the basalts can represent primary zones of secondary porosity. The sequence is expected to behave as a
porosity, but groundwater is predominantly transmitted through the typical mountain block recharge system and to greatly influence alluvial

Fig. 3. - 3D model representing five hydrostratigraphic units, streams and subcatchments. Vertical exaggeration is 12×; vertical extent ranges from −100 m to +1400 m AHD.
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
hydrochemistry, based on field observations and potentiometric sur-
faces (Section 4.1.1). The MRV aquifer features high variability in hy-
draulic properties. For example, reported yields range from b1 L/s up
to 56 L/s (English et al., 2008).
Secondary minerals, such as quartz, fluorite, calcite, clay and zeolite
group minerals, are commonly found in fractures and vesicles of the
MRV basalts, as described in several investigation drill holes in the Too-
woomba area (Free, 1989).
The MRV potentiometric surface indicates groundwater flow from
northeast to southwest in all tributaries, combined with a regional
trend towards the northwest. The potentiometric surface is clearly in-
fluenced by the topography and the presence of the incised alluvial trib-
utary systems, particularly in the two subcatchments to the north (Fig.
4). In these areas, the potentiometric surface suggests the existence of
localised flows towards creeks. This indicates that there is a high likeli-
hood of mixing between alluvial and MRV groundwater as well as po-
tential discharge to surface waters, as previously described by
Martinez et al. (2015). A hydraulic gradient of 0.008 was estimated for
the northern tributaries and 0.012 for Dalrymple Creek.
The Middle Jurassic WCM unconformably underlies the MRV in the
upper parts of the catchment with outcrops in the lower parts of the
tributaries near the main Condamine River channel (Fig. 3). Water bear-
ing zones within the WCM are associated with localised coarser-grained
sandstone and conglomerate beds as well as thin coal seams of b2 m
thickness. In the context of the Surat Basin, the WCM is characterised
as a leaky aquitard for being dominated by low-permeable sediments
like coal, siltstone and mudstone (Smerdon and Ransley, 2012). Howev-
er, when WCM is at shallow depths, under the influence of diffuse re-
charge and higher precipitation rates in the area, usable quantities of
water can be obtained from thin permeable coal seams and sandstones
(Office of Groundwater Impact Assessment, 2016). Such hydraulic
Fig. 4. – Equipotential contour lines of the Main Range Volcanics representing the groundwater elevation pre-1995 and spatial relationship to alluvial aquifer. Equipotential contour line
values are in metres AHD; vertical exaggeration is 12×; vertical extent ranges from −100 m to +1400 m AHD.
7
properties are considered sufficient to promote inflows and mixing
with overlying alluvial groundwaters, where hydraulic gradients are
present.
A pre-1995 potentiometric surface suggests that generally ground-
water flows in a west-northwest direction. A groundwater mound oc-
curs near the boundary between Glengallan and Swan creeks
tributaries (Fig. 1), which is likely to be associated with the Swan
Creek anticline, located immediately to the south. This structure likely
forms a hydraulic barrier to downgradient groundwater flow as this re-
verse fault juxtaposes the WCM against the Marburg Subgroup. A hy-
draulic gradient of 0.005 is estimated for the three tributaries to the
north, increasing to 0.024 in the anticline area.
The upper two members of the Bundamba Group (Gatton Sandstone
and Koukandowie Formation), known as the Marburg Subgroup, underlie
the WCM. The contact between the WCM and Marburg Subgroup is gra-
dational and conformable, and it can sometimes be difficult to identify in
lithological logs. Herein, the contact was identified based on the criteria
suggested by Cameron (1970), who defined this boundary as “the top of
uppermost bed of coarse pebbly sandstone or pebble conglomerate”, indi-
cating that coarse-grained sediments are not expected in the WCM.
The Marburg Subgroup unconformably overlies the basement rocks
of the Texas Beds and intrusive granodiorites that outcrop in the west-
ern flanks of the catchment. The Marburg Subgroup dips and thickens
towards the east, except at the Swan Creek anticline where it outcrops
in a stratigraphic window within the WCM. Marburg Subgroup potenti-
ometric surfaces show that the regional groundwater flow direction is
towards the northwest.
Of the 29 geological units identified to occur in the area, 24
corresponded to Palaeozoic units underlying the western flank of the
Condamine River. The hydrogeological properties of these units were
not investigated due to a lack of lithological and hydraulic data.
8 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18

The uncertainty of the geometry and depth of the layer boundaries
generally increases from the shallower contacts between the alluvium
and underlying bedrock to the contacts between the Marburg Subgroup
and the underlying basement due to the smaller number of bores
intersecting the deeper contacts.
4.1.1. Hydraulic head gradients as indicators of inter-aquifer connectivity
Comparisons between the pre-1995 alluvium potentiometric sur-
face and the hydrogeological model provided insights into possible
zones of connectivity and mixing (Fig. 5 a, b, c, d). Strong evidence of
hydraulic mixing between the MRV and the alluvium was identified
by comparing the pre-1995 potentiometric surface to the 3D extent of
the alluvium, and to the alluvium potentiometric surface. The equipo-
tential lines of the MRV indicate flow towards the creeks in most tribu-
taries (Fig. 4), and the head gradient maps indicate that there is a
gradient from the MRV to the alluvium in the majority of the area
where data exist. This suggests that the MRV are a major source of in-
flow to the alluvium. Connectivity may exist particularly along transi-
tion zones with higher permeability or at the breaks of slope of the
topographic surface. Preferential flow paths are likely to be present in

Fig. 5. – Hydraulic gradient maps showing the spatial distribution of relative contribution of four end-members, as discussed in Section 4.6: a) MRV-WCM; b) alluvium-MRV; c) alluvium-
Marburg Subgroup; and d) alluvium-WCM. Green values indicate hydraulic head in the overlying unit is higher than the underlying (corresponding to a downwards hydraulic gradient).
The longitudinal transect projected in d) is presented in Fig. 12 (Section 4.6.1). The extent of the gradient maps is limited by the data availability footprint for the two units assessed in each
map.
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
zones of alignments of fractures, faults and joints, as suggested by sharp
bends of creek lines notably present in the upper sections of Hodgson
and Kings creeks (Fig. 5 a).
The spatial distribution of upward hydraulic gradients from the
WCM (Fig. 5 d) is similar to that of the MRV (Fig. 5 b). Confluences of po-
tentiometric contours near creek lines of two tributaries to the north
suggest that the alluvium also receives groundwater discharge from
this coal-bearing unit. Consequently, WCM groundwaters were not
discarded as a potential recharge source to the alluvium.
A similar spatial distribution of upward gradients is also observed for
the Marburg Subgroup (Fig. 5 c), predominantly in the lower sections of
the tributaries and proximal to the Marburg Subgroup outcrop area.
Some degree of connectivity between WCM and MRV is also expect-
ed based on comparisons between potentiometric surfaces. These sug-
gest that downward fluxes from the MRV to the WCM may occur near
the catchment boundary of Hodgson and Kings creeks. Downward
fluxes may also occur towards the eastern boundary of the study area
near the Great Dividing Range. Here the WCM dips towards the east
and the MRV becomes thicker, thereby increasing the potential for
downward flux due to an increased hydraulic gradient. Elsewhere, hy-
draulic gradients indicate potential areas of upward flux from the
WCM to the MRV (Fig. 5 a).
It is important to note that these potentiometric surfaces were de-
rived from hydraulic head observations from the period 1940–1995.
This period was prior to recent persistent droughts (1995–2009), dur-
ing which maximum groundwater extraction occurred (van Dijk et al.,
2013). Furthermore, the interpolation of data points based on kriging
is subject to a degree of uncertainty, which was assessed by the spatial
distribution of variance maps to identify zones where the results of
such interpretations should be taken with more caution.
4.2. Hierarchical cluster analysis
Seven hydrochemical water groups were defined via Hierarchical
Cluster Analysis (HCA) with a summary of the median values and stan-
dard deviation of each variable according to the clusters presented in
Appendix C. This HCA builds upon the work by Martinez et al. (2015),
which presents a thorough description of each water type, their spatial
distribution and relation to the different hydrogeological units. The
main difference between the current HCA outputs and those provided
by Martinez et al. (2015) is the recognition of an additional group (clus-
ter 5), resulting from the inclusion of rainfall data in the analysis and the
division of surface water hydrochemical data into two seasons: dry and
wet, as per seasonal patterns described in Section 2.1. The combination
of HCA with PCA helped to refine the understanding on inter-aquifer
connectivity and a discussion is presented in Section 4.3.1. The spatial
distribution of the hydrochemical water groups in a 3D perspective re-
lated to the alluvium body is shown in Fig. 6.
4.3. Principal component analysis
To develop an understanding of hydrochemical processes control-
ling rainfall, groundwater and surface water (seasonally), Principal
Component Analysis (PCA) was conducted and the results were put in
the context of the HCA-derived water chemistry groups and represent-
ed using the 3D model. Furthermore, for the identification of prelimi-
nary end-members, five tests were conducted using different
combinations of chemical variables via PCA based on the same dataset
interpreted via HCA. Test #1 had 994 complete cases considering all
12 variables referred in Section 3.4, with a combined variance for
three components of 64.5%. The variables F, SiO2 and pH showed
weights below 0.18 for both PC1 and PC2·This guided the combination
of variables used for tests #2 to #5, either by removing one or all of these
three variables in different combinations.
The output of test #4, which excludes SiO2 and pH (Fig. 7), had a
combination of variables explaining the highest variance in the dataset
9
with weights of approximately ±0.3 or higher. Three principal compo-
nents (PCs) with eigenvalues greater than one (Kaiser, 1958) explain
together 74.3% of the variance in the dataset.
PC1 is responsible for 47.8% of the total variance and is characterised
by positive weighting of all variables (Fig. 7). Variables with the highest
loadings are EC (0.44), Cl (0.41) and Na (0.37). Samples with high scores
of this component are likely to have been largely influenced by
weathering and evapotranspiration processes. Samples with PC1 values
greater than two are located primarily at the lower parts of the tribu-
taries in both sides of the Condamine River and mostly away from
major tributaries.
PC2 explains 16.3% of the total variance and is characterised by pos-
itive weightings for Ca, Mg, and HCO3 and negative scores for Na, K and
SO4. This may indicate that cation exchange with clay minerals is likely
to occur via replacement of Ca and Mg by Na and K. GW samples with
−2.0 b PC2 N 1.8 confirm that such processes are likely to occur through
the whole extent of the alluvium in the Dalrymple and Glengallan
creeks and within the MRV in the upper parts of the northern tributaries
(Hodgson and Kings creeks) (Fig. 1). Similar patterns are also observed
in WCM and Marburg Subgroup samples throughout the whole extent
of their surface outcrops.
Strong positive loading for F and negative for K and to a lesser extent
Ca characterise PC3. This PC represents only approximately 10.2% of the
variance in the dataset and samples from all major HSUs with PC values
N2 do not show any spatial pattern or association to a particular
hydrostratigraphic unit.
4.3.1. Relationship between PCA and hydrochemical facies
Bivariate plots of PC1 and PC2 loadings for all samples are shown in
Fig. 8, with colouration by (a) cluster and (b) hydrostratigraphic unit.
These two components explain together approximately 64% of the var-
iance in the dataset and the positive score of samples indicate that salin-
ity and possible cation exchange are directly dependent/or correlated to
the hydrogeological context and hydrochemical facies they are related
to.
Samples of cluster 1 correspond to Mg-HCO3 type water with ‘brack-
ish’ salinity levels (median EC ~ 1500 μS/cm). They are somewhat influ-
enced by PC 1 and to a higher extent by PC 2, with the majority of the
samples positioned at the top right quadrant of the biplot (Fig. 8 a).
This group is dominated by MRV samples (55%) and overlaps with sam-
ples of clusters 2, 3 and 6, suggesting that mixing among these groups
may occur.
The great majority of samples assigned to cluster 2 have positive
loadings for component 2 and negative loadings for 1. Most of these
samples are fresh surface water samples (SW_D, corresponding to 46%
among all SW) and groundwater samples (65% alluvium) of Mg-HCO3
water type. Most groundwater samples in this group were collected
from alluvial aquifers in Glengallan Creek (Fig. 6).
Samples assigned to cluster 3 are positioned in the centre of the bi-
variate plot, with no clear distinction from other clusters. Alluvial sam-
ples correspond to 37% of all 175 cases, followed by MRV (31%). This
group is characterised by fresh to brackish (median EC 1135 μS/cm)
Na-HCO3 type water. The chemical characteristics of these samples
strongly suggest that they may result from mixing between different
groundwater sources in MRV and WCM units (predominantly clusters
1 and 4).
With the second lowest salinity values among all clusters, samples of
cluster 4 have both negative loadings for PC1 and PC2 and are dominat-
ed by MRV samples (81%), followed by WCM (14%), with Na-HCO3,(Cl)
water type. This hydrochemical signature suggests this group as a po-
tential MRV end-member. Furthermore, the fact that this cluster is not
composed of samples from an unique hydrostratigraphic unit and that
the samples are the deepest samples among all clusters (median
depth of 65 m), MRV groundwaters likely mixes with WCM and the
end-member may not necessarily represent an individual unit but to a
zone of interaction between these two formations (WCM-MRV). For
10 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
Fig. 6. – Three-dimensional distribution of seven clusters (HCA) superimposed onto the 3D geological model. Vertical exaggeration is 12×; vertical extent ranges from −100 m to
+1400 m AHD.
the purpose of clarity in the discussions around mixing calculations, we
refer to this group of groundwaters as the WCM end-member.
Strong negative loadings of PC's 1 and 2 characterise samples
assigned to cluster 5. This group contains the freshest samples of all
clusters (~190 μS/cm), and most samples (86%) assigned to this cluster
correspond to rainfall or surface waters (36% SW wet and 36% SW dry).
This group clearly corresponds to recharge from the land surface direct-
ly on to the alluvial aquifer.
Samples assigned to cluster 6 are characterised predominantly
by positive loadings of PC1 and negative loadings of PC2. Water
samples in this cluster have a Na,(Mg)-Cl type with the second
highest median EC among all groups (~ 2575 μS/cm). This group is
composed mainly of groundwater samples collected from the allu-
vium (45%), WCM (24%) and MRV (17%). Overlaps with other
groups (3 and 7) suggest that alluvial samples from this group
may be influenced by some degree of mixing with WCM and/or
Marburg Subgroup.
Among all groups, samples assigned to cluster 7 represent the most
distinct water type (Na-Cl), characterised by high positive values of PC1
(median EC = 6622 μS/cm) and negative loadings of PC2. Apart from al-
luvial samples, most of cluster 7 related samples is sourced from
Marburg Subgroup (36%, Fig. 8 b), strongly suggesting that this group
represents a possible MS end-member.
The level of heterogeneity recognised in the lithological profiles
of hundreds of bores drilled into the MRV and used for the develop-
ment of the 3D geological model was confirmed by identification of
up to five different water types (1, 2, 3, 4 and 6), with median EC
values ranging between 624 μS/cm and 2575 μS/cm. The different
MRV water types are associated with different depth intervals and
confirm the high level of heterogeneity in this unit. Moreover, com-
parisons between spatial distributions of hydraulic head gradients
and hydrochemical composition indicated that deeper parts of the
MRV are featuring some level of connectivity to the underlying
Walloon Coal Measures (WCM).
4.4. Tracers selection and identification of preliminary end-members
The preliminary assessment of the suitability of different combina-
tions of tracers to identify sources of inflow to the alluvial aquifer
based on five tests (Section 4.3) showed that exclusion of SiO2 and pH
from an initial group of 12 species resulted in the highest cumulative ex-
plained variance from three PCs in test#4 (Fig. 7). Following the exclu-
sion of SiO2 and pH, the set of ten variables was then used for
identification of preliminary end-members (Fig. 8b), and helped guid-
ing the collection of additional field and laboratory data for better char-
acterisation of these prospective end-members.
Based on this identified set of tracers, the hydrochemical data from
the analytical program carried out in mid-2014 and mid-2015 were
combined to determine the most adequate species for the mixing calcu-
lations. The reason for selecting the smaller dataset in the mixing calcu-
lations was that all samples were collected under similar meteorological
conditions (dry) and within a short time span, reducing the uncer-
tainties associated with temporal variation in end-members and with
additional support from the preliminary EMMA findings. More conser-
vative species such as stable isotopes of water (δ18O) were incorporated
into the set of variables. Furthermore, residual alkalinity (RA) which
corresponds to alkalinity + [F]-[Ca], where [F] and [Ca] are the total con-
centrations of F and Ca (that is, including ionic species and all their com-
pounds) was included as it is assumed to be unaltered by calcite or
fluorite chemical dissolution or precipitation (Barbiéro et al., 2001).
The set of variables selected for mixing calculations corresponded to
EC, δ18O, Na, K, Ca, Mg, HCO3, Cl, SO4, F and RA. By adding δ18O and RA
and removing NO3, approximately 81% of the variability in this dataset
could be explained by three eigenvectors. Anthropogenic sources of
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
Fig. 7. – Three principal components (PC) obtained from Principal Component Analysis of
historical groundwater, surface water and rainfall data. The sum of the variance from the
three principal components is shown in brackets for each test.
NO3 are unevenly spread across the region (Hansen, 1999), which could
skew its representation in the dataset, associated with potential denitri-
fication processes possibly enhanced by mixing of deeper and reduced
groundwater with more oxidised alluvial waters.
Fig. 8. – plot of loadings of components 1 and 2 with samples identified by a) cluster group and b) hydrostratigraphic unit and water type (SW). Samples at the vertices of the polygon were
identified as preliminary end-members via EMMA.
11
According to the ‘rule of 1’ (Hooper, 2003), an eigenvector (n) must
explain N 9.1% of the variance for a sample containing eleven species [1/
number of species]. For this particular dataset, eigenvector 1 explained
54.7% of the data variability, eigenvector 2 explained 15.6% and eigen-
vector 3 represented 10.3% of the variability.
The results from the Bartlett's sphericity test (chi-square statistic =
2388; degree of freedom = 55; p b 0.001) confirm that there is common
variance shared among the selected variables. The KMO test result of
0.669 also indicates that there is inter-correlation between the vari-
ables. This confirms that PCA is useful for reducing the dimensionality
of the dataset and can be applied for the EMMA.
4.5. End-members characterisation
Huxley (1982) proposed that alluvial hydrochemistry was influ-
enced by recharge from streams and inflow from adjacent bedrock aqui-
fers. As a result, the author identified two end-members: a mix of MRV
with stream waters and groundwaters associated with WCM and Mar-
burg Subgroup. The study by Huxley (1982) focused on the central
and broader parts of the Condamine River alluvial system, mostly locat-
ed outside (further downgradient) of the study area. However, the dy-
namics and conceptual framework are applicable to the headwaters of
the catchment as well.
Huxley and QWRC (1982) further suggested that rainfall is the pri-
mary source of water for the alluvial aquifer system, recharging indi-
rectly via bedrock formations and/or via surface water, particularly in
areas where the streams are influent. Rainfall is expected to significantly
contribute to recharge of the alluvium at the floodplains of mid to lower
parts of tributaries and along the Condamine River system. This is sup-
ported by recent findings of Hocking and Kelly (2016) on a regional
scale and localised estimations to the north of the study area (Owen
and Cox, 2015). Rainfall can be considered as clean natural recharge
water with samples collected for the period between January 2014
and June 2015 (Appendix B). Sample RF02B (Fig. 1) represents the re-
charge pattern throughout most of the study area outside of the influ-
ence of the orographic effect created near the Great Dividing Range.
Other relevant alluvial recharge sources include contributions from
underlying aquifers, particularly associated with different water bearing
zones of MRV. The indication of possible mixing processes between
these units and the alluvium is strongly supported by the hydraulic gra-
dient maps (Fig. 5 a, b, c, d). Bore P0768A was identified as a represen-
tative example of MRV-cluster 2 groundwaters (EM-B) which have the
12 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
lowest median sample depth among the five MRV-related clusters
(~20 m bgs), as discussed in Section 4.3.1.
Other MRV associated end-members were also considered in a sen-
sitivity analysis due to their distinct characteristics demonstrated by
EMMA (clusters 1 and 4, respectively). However, these samples corre-
spond to possible mixing between MRV cluster 2 with Marburg Sub-
group and with rainfall and WCM, respectively. There was no
indication from other existing lines of evidence confirming them as pro-
spective end-members for the mixing calculations.
The underlying sedimentary bedrock units of WCM and Marburg
Subgroup were also identified as likely end-members in areas where
upward hydraulic gradients indicate possible connectivity with the allu-
vium, as also supported by isotope signatures discussed in Martinez
et al. (2015). Sample P1222A, from a depth of 97.70 m below ground
surface (bgs) was considered representative of WCM groundwaters
(EM-D) for the mixing calculations discussed in the following section.
Bore PRV05A2 corresponds to the end-point of a cluster with elevat-
ed positive weights of PC1 (median EC ~ 6600 μS/cm) and negative PC2
values (EMMA), indicating predominance of Na over Ca and Mg and Cl
over HCO3. This location is considered representative of Marburg Sub-
group (end-member EM-C) due to its location in the centre of one of
its outcropping zones and standing water level at approximately 45 m
bgs.
Surface water samples collected during high flow conditions (‘SW
wet’) were also initially tested as possible EMs. These samples are
characterised by negative PC1 values and they are positioned between
two clusters of samples comprising rainfall and MRV-cluster 2 (Fig. 8
b), composed of the shallowest MRV sample depths (median depth of
~20 m bgs) located predominantly in the tributaries to the north. The
EMMA suggests that these shallower MRV aquifers are replenished
during higher precipitation events, and commence to discharge into ad-
jacent alluvium and surface water following short time lags, contribut-
ing to mixing in these water compartments. This hypothesis is
corroborated by the findings of Martinez et al. (2015), based on stable
isotopes and 222Rn measurements. As a consequence, ‘SW wet’ was ex-
cluded as a potential EM. Moreover, surface waters collected during
baseflow conditions (‘SW dry’) do not present a distinct signature in
EMMA (Fig. 8 b), as the majority of the samples are located closer to
the centre of the biplot. The overlapping with areas occupied by alluvi-
um and MRV-cluster 2 samples confirms the contribution of MRV and
alluvial groundwater towards creek water chemistry during baseflow
conditions. Similarly, SW dry samples were then included as mixed
samples in the mixing calculations rather than as end-member.
Considering the above scenario and according to the rule of n + 1,
where n = principal components with eigenvalues N1 (Joreskog et al.,
1976), a minimum of four end-members is required to explain the
mixing effect in the geochemical composition of these samples. Conse-
quently, four samples were chosen (Table 1) as representative end-
members of possible sources of recharge in the region in the context
of the hydrogeological conceptual model framework described in previ-
ous sections. Their association with different cluster groups (water
types or hydrochemical facies) strongly supports the choice of these
samples as representative end points in the water chemistry balance
of this upper catchment.
Table 1
– Hydrochemical and isotopic properties of selected end-members used in the mixing computations.
Selected Cluster δ18O Deuterium
EM Sample Source group (‰) (‰)
A RF02Bb Meteoric 5 −4.17 −20.00
B P0768B MRV 2 −4.18 −26.07
C PRV05A2 Marburg 7 −5.09 −30.51
D P1222A WCM 4 −4.99 −28.61
a
– a complete list of saturation indexes (SI) for 14 minerals calculated using PHREEQC (Parkhurst and Appelo, 2013) is shown in Appendix B.
b
– stable isotopes values correspond to precipitation weighted for Brisbane (Hughes and Crawford, 2012).
4.6. Sources of alluvial aquifer inflow and spatial distribution of mixing
ratios
The mixing calculations allowed estimation of the contribution ra-
tios for both mixed alluvial groundwaters and surface water samples,
which were interpreted in the context of the hydrogeological conceptu-
al model.
PCA of the mid-2014 and mid-2015 dataset (Fig. 9 a) allowed the es-
timation of the range of variance, covering the proportion that may be a
result of unrecognised end-members or related to non-mixing process
(e.g., water-rock interaction). By using these projections, an iterative
process of end-member and species selection was carried out until an
acceptable conversion of mixing was achieved by assessing the total
and individual contribution of species and end-members to the objec-
tive function.
Standard deviation (σ) values were assigned to both end-members
and mixed samples considering the uncertainties already described in
Section 3.5. For all end-members, standard deviation values of 100
times their respective sample concentration were used. For the mixed
samples, lower σ values, between 0.7 and 3.0 times the mean concen-
tration of each parameter, were assigned, depending on their behaviour
(conservative vs non-conservative) and origin (i.e., groundwater, sur-
face water and rainfall).
Following the calculation of mixing ratios using the estimated con-
centration of end-members, a number of surface water and few alluvial
samples still deviated from the 1:1 mixing line between measured and
estimated values for each species (Fig. 10). The results indicate that
only EC, Na and Cl behave conservatively, whereas δ18O, Ca, Mg, K,
HCO3, SO4 and F do not. The non-conservative behaviour of δ18O is ex-
pected by evaporation fractionation affecting surface water and shallow
alluvial samples. The non-conservative behaviour of other species such
as Ca, Mg, K, HCO3 and SO4 is attributed to geochemical reactions as
discussed in the following section.
The spatial distribution of the relative contribution of four end-
members (A to D) exposed correlations between the sources with par-
ticular hydrostratigraphic units and/or geomorphological features
(Fig. 5 a, b, c, d).
The meteoric source (end-member A) contributes to 42% of overall
recharge among all samples, with a lower ratio (35%) between the 47 al-
luvial mixed samples. The groundwater samples with end-member
ratio higher than 50% are preferentially positioned in the lower parts
of the tributaries and more predominantly along the wider channel of
the Condamine River (Fig. 5 a). End-member A contributes to approxi-
mately 57% among surface water samples, reaching up to 97%
(SW24_D).
The MRV related end-member B has the highest contribution ratio
(41%) among the alluvial groundwater samples, with a lower value
when assessing all mixed samples combined (38%). High contribution
rates (N50%) of end-member B are noted at the north-eastern and cen-
tral eastern parts of the area, where thicker sequences of these volcanic
rocks and steeper hydraulic gradients occur (Fig. 5 b). In alluvial
groundwater, approximately 50% of the samples had end-member B
participating with N50% of the mixture, indicating the importance of
this source in the groundwater balance of this shallow aquifer system.
EC Sample depth (m Elevation (m Aragonite Calcite Dolomite
(μS/cm) bgs) AHD) SIa SIa SIa
29 0 473.4 −3.0 −2.8 −5.8
994 14.5 520.6 0.4 0.5 2.2
8730 44 456.6 −3.5 −3.3 −3.0
547 95.5 498.8 −1.6 −1.4 −3.0
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18 13

Fig. 9. – Bivariate plot of PCs (PC1, PC2) of (a) 2014–2015 data set (n = 97) and (b) relative component weights of the estimation from the mixing calculations. Note: end-members (EM)
are identified with a black square and alluvial with a red circle.

Fig. 10. – Bivariate plots of measured versus estimated concentrations of the eleven species considered in the mixing calculations. The four end-members are identified with respective
letters in red colour.
14 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
The end-member B ratio in surface water samples can reach up to
95%, confirming the findings of Martinez et al. (2015) and suggesting
that the hydrochemistry of surface water is controlled mainly by mixing
between end-members A and B, with certain WCM participation in their
outcropping areas.
Overall, the participation of the Marburg Subgroup (end-member
C) in hydrochemical mixing is relatively small (5% and 7% in all samples
and alluvium, respectively), and it is restricted to areas where the allu-
vium directly overlies Marburg Subgroup. The highest values of end-
member C input into the alluvium (up to 66%) are found in the north-
west corner of the study area (Fig. 5 c), where relatively thin alluvium
directly overlies the Marburg Subgroup, under strong evapotranspira-
tion influence (Martinez et al., 2015), affecting groundwater of both
hydrogeological units. Moreover, the intense groundwater extraction
from the shallower aquifer may induce flow from the underlying
bedrock.
Finally, WCM source (end-member D) showed 14% mean contribu-
tion among all samples, with higher participation in alluvial
hydrochemistry (17%). Zones where higher contribution rates (N 30%)
were observed coincide with the outcrop areas of the WCM near the
main Condamine River channel (Fig. 5 d), where hydraulic gradients
are favourable for this occurrence.
Except for P0768B, the representative sample of end-member B,
other ten samples from bores targeting MRV aquifer had some contribu-
tion of end-members A (rainfall) and D (WCM), reaching up to 80%
(12% in average). This indicates that MRV is replenished by rainfall,
and more importantly that some degree of mixing with WCM is likely
to occur, a finding corroborated by the hydraulic gradient map (Fig. 5 a).
The four end-member proportion ratios in 44 alluvial samples and
the respective cluster groups are shown in Fig. 11 and the results for
all samples are presented in Appendix B. The analysis of recharge contri-
bution ratios and cluster membership confirms that end-members A
and B are the major recharge sources to alluvial groundwaters assigned
to clusters 1 and 2 (relatively fresh, MgA\\HCO3 water types), whereas
Fig. 11. – Contribution ratios in alluvial mixed samples according to cluster group. Notes: EM-A: rainfall; EM-B: MRV groundwaters, EM-C: MS groundwaters and EM-D: WCM
groundwaters.
increased ratios of end-members C and D are found in samples assigned
to clusters 3 (Na\\HCO3), 6 (Na\\Mg\\Cl) and 7 (Na\\Cl), denoting an
evolution in hydrochemical signatures.
4.6.1. Longitudinal variation in end-member contribution ratios to alluvial
aquifer inflows and hydrochemical evolution
When assessing the longitudinal variation of recharge ratios to
alluvial groundwaters across a transect of a representative tributary
(Dalrymple Creek) where all four recharge sources are expected to in-
fluence in alluvial chemical composition (Fig. 12), the importance of hy-
draulic gradients and physical aquifer properties (i.e., aquifer
thicknesses, interfaces and permeability distribution) becomes more
evident. According to the mixing calculations, in the upper part of the
tributary, where the alluvium is very thin and composed mainly of
basalt-related boulders and gravels, end-member B (MRV) contribution
to alluvial recharge is the highest (up to 75%) and the contribution from
the WCM's is on average approximately 9%. In this upper part of the trib-
utary, alluvial recharge sources are comprised predominantly of a mix
of rainfall, MRV and WCM end-members, with predominance of the
first two. Towards mid-subcatchment, an increased contribution of
WCM and locally Marburg Subgroup is identified, coinciding with
favourable hydraulic gradients (WCM N alluvium) and less influence
of MRV hydraulic load on alluvial hydraulic dynamics.
Among the eight surface water samples collected along this creek,
end-member A (rainfall) responds to 68% of mixing ratio in average,
whereas MRV has a mean contribution value of 32% (up to 70%). It is
valid to note that surface water samples were collected following a
prolonged dry period, suggesting that they are resulted from meteoric
recharge of MRV and alluvium in the higher parts of these
subcatchments with subsequent discharge following a short residence
time, in zones of topographic break (within 40–50 km range). These
stream waters, characterised by signatures of end-members A and B
continue to flow downstream at the surface and finally recharge the al-
luvium in the lower floodplains, where the creeks are predominantly
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18 15

Fig. 12. – Longitudinal section of Dalrymple Creek showing the influence of geology and hydraulic gradients on the spatial distribution of recharge sources contribution to alluvium
groundwater and surface water samples. End-member samples: P0768B, G90 (P1222A), G49 (PRV5-2A) and alluvial groundwater samples P0701A and P1331B were used in the
geochemical evolution assessment of flow path FP1.

influent, as alluvial groundwater levels are considerably below the base where Andes. –andesine; Dolom.–dolomite; Kaol.–kaolinite; gw_up –
of the creek (Martinez et al., 2015). This can be noted on Fig. 12, which groundwater composition at initial sample; gw_down –groundwater
shows a reduction in ratios of these two end-member in the mid- composition at final sample; MRVgw and WCMgw – groundwater end-
tributary (within 20–30 km range) and returning to higher values to- members involved in mixing (percentage). The values shown next to
wards downstream (0–15 km range), near the confluence with main the mineral phases correspond to mmol/kg of water, evaporation factor
Condamine River. next H2O(g) ↑ is in kilograms of water.
Some reactive species (Ca, Mg, K, HCO3 and SO4) used in the mixing The criteria for selection of the most plausible model in segments 1a
calculations diverged from the 1:1 mixing line, indicating their partici- and 1b was based on saturation indexes of initial and final samples, ion
pation in chemical reactions also controlling the evolution of these wa- ratios that are indicative of predominant weathering processes within
ters. The fact that δ18O also did not behave conservatively reflects the these particular segments and the relative mixing values previously ob-
influence of evapotranspiration on a group of samples. The relevance tained via linear mixing calculations.
of this process should also be confirmed independently by other lines The evolution of alluvial hydrochemistry confirms that various
of evidence. For that reason, an analysis of the geochemical evolution water-rock processes contribute to the steady increase of TDS values
of alluvial groundwaters along an inferred flowpath (Fig. 5 d) was car-
ried out with the purpose of characterising the predominant rock-
water reactions associated with this aquifer. Table 2
– Data of two segments of the flowpath identified for inverse model, including data of
Table 2 summarises the data and characteristics of two segments of
mass transfer (mmol/kg of water) for identified net chemical reactions.
flowpath 1 (FP-1a and FP-1b), including the description of predominant
and more likely processes controlling groundwater evolution. A sum- Flowpath 1 Segments
mary of the constraints (dissolved elements) and phases applied in 1a 1b
the models is presented in Appendix D. Initial sample RF02B* P0701A
The following net balance equations were achieved for each of the Final sample P0701A P1331B
selected segments (Fig. 12): [TDS, in mg/L] [394.2] [961.8]
(water type) (Mg,(Ca)-HCO3) (Mg,(Na)-HCO3)
Geological Upper tributary, thin alluvium Mid-tributary, alluvium
Segment1a : 65%Rainwater þ 35%MRVgw þ 0:35Andes: domain at interface with and under overlying contact surface
þ 0:08Olivine þ 0:26Augite influence of MRV between units MRV and WCM
þ 0:54CO2ðgÞ → Alluvgwd own þ 0:24Kaol: þ 0:03Illite ð4Þ EMs Meteoric (RF02B): 64 Meteoric (RF02B):26
contribution MRV (PRV10A): 36 MRV (PRV10A): 59
ratios (%) via WCM (P1222A): 15
mixing only
Major physical Mixing, basalt minerals Mixing, plagioclase and
and chemical dissolution, CO2(g) uptake, and dolomite dissolution, ion
Segment1b : 10%Alluvgwu p þ 71%MRVgw þ 19%WCMgw processes precipitation of clay minerals exchange, clay minerals
þ 0:71Dolom: þ 1:12Na2 X þ 0:32Andes: → Alluvgwd own precipitation and evaporation
þ 1:12CaX þ 0:22Kaol: þ 0:02Illite þ 1H2 OðgÞ ↑ ð5Þ Notes: * RF02B - TDS: 17.7 mg/L; Na,(Mg)-HCO3,(Cl) water type.
16 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
towards downstream and confirms the relevance of mixing to spatial
variation of chemical signatures, via contribution of up to three end-
members into recharge.
In the upper parts of the tributary system (segment 1a), the dissolu-
tion of primary minerals associated with MRV basalts (i.e. olivine, au-
gite, and andesine) via recharge of meteoric waters is the major factor
releasing the elements found in the groundwater of the alluvial aquifer.
In this part, the alluvium is predominantly composed of boulders and
large gravels of basalt components eroded and transported over short
distances from the steep MRV outcropping zone located immediately
upstream. The precipitation of kaolinite and illite is possibly associated
with andesine weathering.
As the groundwater evolves towards the flatter zones of the tribu-
tary system (segment 1b, approximately 20 km downstream), fresh ol-
ivine is not expected to be present in the aquifer matrix anymore.
However, dissolution of andesine combined with ion exchange and do-
lomite dissolution can release Ca, Mg, Al and Si into the local groundwa-
ters. A larger proportion of MRV contribution compared to the upper
segment occurs followed by possible mixing with WCM groundwaters.
The minerals calcite and kaolinite are expected to precipitate towards
the downstream tributary section. Considering that the alluvium be-
comes wider in an area of higher evapotranspiration potential, such
processes were included in the simulation, with an estimated evapora-
tion factor of 1 obtained through the inverse modelling. This ultimately
supports the findings of Martinez et al. (2015), particularly in relation to
isotope signatures of these samples, which created an evaporation line
δ2H = 4.3 δ18O–5.5 when combined with surface water samples data.
5. Conclusions
Hydrochemical mixing models are subject to considerable uncer-
tainties, as they require detailed knowledge of zones where the differ-
ent components of hydrological systems interact. Due to the
complexity and data sparseness of many hydrological systems, the ap-
plication of hydrochemical mixing approaches without an adequate un-
derstanding of actual mixing pathways or approaches based on singular
lines of evidence are likely to provide unreliable recharge estimates. To
overcome such limitations, innovative approaches that integrate multi-
ple independent lines of evidence are required.
This study shows how the development of a robust conceptual
hydrogeological model based on physical, hydraulic and hydrochemical
data prior to using end-members mixing models can provide reliable
estimates of the relative contribution of alluvial recharge sources from
multiple bedrock aquifer systems. Although these physical, hydraulic
and hydrochemical techniques are well-documented, the integrated
and iterative manner of how these techniques are combined in this
study is novel, resulting in increased confidence in selecting
hydrochemical end-members and constraining sources of inflow to
the alluvial aquifer.
As part of a sequential workflow, we have first developed a three-
dimensional (3D) geological model of the headwaters of the Condamine
River basin (Queensland, Australia) to consider where inter-aquifer
mixing is likely to occur based on the physical attributes of the aquifer
system (e.g. aquifer geometry). The model was then combined with po-
tentiometric surfaces, hydrochemical data and hydraulic head gradient
maps, providing further evidence to support identification of zones
where mixing between aquifers occurs.
The Main Range Volcanics (MRV), a major regional-scale fractured
rock aquifer system with an average thickness of over a hundred metres,
underlies the alluvium system throughout most the upper reaches of
the catchment. Mixing calculations estimated an average inflow contri-
bution from the MRV to the alluvium of over 40%, which is higher than
the estimated average diffuse recharge (35%) and contribution from
two adjacent sedimentary bedrock aquifer systems (combined contri-
bution of 24%).
The spatial assessment of hydrochemical evolution along inferred
flow paths within a sub-catchment of the Condamine River basin con-
firms that mountain block recharge has a significant influence on the
variability of hydrochemistry within the alluvial aquifer. Interestingly,
this assessment also shows that the influence of mountain block re-
charge on the hydrochemistry of the alluvial aquifer system extends
considerable distances downgradient from the catchment headwater
zone, where the majority of precipitation and localised recharge occurs.
This highlights that there is a need to account for these sources in water
balance studies and to assess their influence on hydrochemical evolu-
tion along inferred groundwater flow paths.
As the reliability and value of recharge source mixing calculations
strongly depends on a realistic understanding of actual flow paths and
interfaces between different aquifers or between groundwater and sur-
face water, a similarly integrated and sequential approach that takes
into account multiple independent lines of evidence is recommended
for other studies.
Although this approach was in this study applied to a headwater
catchment, the steps adopted here can be applied to investigate inter-
aquifer recharge, water quality issues and recharge processes across a
wide range of hydrological and climatic settings. For example, the meth-
od can assist in identifying potential risks of contamination to shallower
aquifers due to mixing with underlying bedrock units that may contain
lower-quality groundwaters such as saline connate water, or it may
help to detect the presence of dissolved pollution plumes associated
with anthropogenic activities. It can also assist in building the model
structure for investigations of potential negative effects to storage, hy-
draulic levels and groundwater quality of shallower aquifers and conse-
quently surface water bodies as a result of depressurisation of
underlying hydrostratigraphic units associated with onshore oil and
gas production activities.
Acknowledgments
The authors would like to thank the assistance of S. Russel, A.
Bhargav and C. Allen, from the Central Analytical Research Facility
(QUT) for the valuable assistance with the analytical procedures. The
field and laboratory activities would not have been possible without
the financial contribution received and the infrastructure provided by
the School of Earth, Environmental and Biological Sciences, facilitated
by M.E. Cox (QUT), who is greatly thanked. The authors also appreciate
the comments received from A. Mckay, L. Reading and the review by C.
Turnadge and M. Gilfedder (both CSIRO Land and Water). We acknowl-
edge the application of Leapfrog™ Geo software, student license in this
paper and express gratitude to A. Jurado for the comments on the
mixing calculations and D. Owen for the support with saturation index
calculations. The authors wish to thank the Co-editor-in-Chief D. Barceló
Cullerés and four anonymous reviewers, whose comments helped to
improve the manuscript. The first author greatly appreciates the finan-
cial support provided by an Australian Postgraduate Award scholarship
and top-up scholarships by QUT and CSIRO Land and Water.
Map. KML file containing the Google map of the most important
areas described in this article.
Appendix A. Supplementary data
Supplementary data associated with this article can be found in the
online version, at http://dx.doi.org/10.1016/j.scitotenv.2016.09.029.
These data include the Google map of the most important areas de-
scribed in this article.
References
Aishlin, P., McNamara, J.P., 2011. Bedrock infiltration and mountain block recharge ac-
counting using chloride mass balance. Hydrol. Process. 25, 1934–1948.
 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
Barbiéro, L., Valles, V., Régeard, A., Cheverry, C., 2001. Residual alkalinity as tracer to esti-
mate the changes induced by forage cultivation in a non-saline irrigated sodic soil.
Agric. Water Manag. 50, 229–241.
Bureau of Meteorology, 2015. Climate Data Online. 2014.
Cameron, J., 1970. The Rosewood-Walloon Coalfield. Vol 344. SG Reid.
Carrera, J., Vázquez-Suñé, E., Castillo, O., Sánchez-Vila, X., 2004. A methodology to
compute mixing ratios with uncertain end-members. Water Resour. Res. 40,
W12101.
Cartwright, I., Weaver, T., Cendón, D.I., Swane, I., 2010. Environmental isotopes as indica-
tors of inter-aquifer mixing, Wimmera region, Murray Basin, Southeast Australia.
Chem. Geol. 277, 214–226.
Cendón, D.I., Larsen, J.R., Jones, B.G., Nanson, G.C., Rickleman, D., Hankin, S.I., et al., 2010.
Freshwater recharge into a shallow saline groundwater system, Cooper Creek flood-
plain, Queensland, Australia. J. Hydrol. 392, 150–163.
Christophersen, N., Neal, C., Hooper, R.P., Vogt, R.D., Andersen, S., 1990. Modelling
streamwater chemistry as a mixture of soilwater end-members — a step towards
second-generation acidification models. J. Hydrol. 116, 307–320.
De Gaspari, F., 2015. Mixing and speciation algorithms for geochemical and reactive
transport problems. Materia (s) 30, 04–2015.
DNRM, 2013a. Queensland Groundwater Database.
DNRM, 2013b. In: DNRM (Ed.), Surface Water Database.
Doctor, D., Alexander Jr., E.C., Petrič, M., Kogovšek, J., Urbanc, J., Lojen, S., et al., 2006.
Quantification of karst aquifer discharge components during storm events through
end-member mixing analysis using natural chemistry and stable isotopes as tracers.
Hydrogeol. J. 14, 1171–1191.
English, P., Levett, J., Lewis, S., Sandow, J., 2008. 3D Geologic Modelling for Groundwater
Applications in the Condamine Headwater Sub-catchments, Queensland. Geoscience
Australia, Canberra, p. 45.
Field, A., 2009. Discovering Statistics Using SPSS. Sage publications.
Fielding, C., 1993. The middle Jurassic Walloon coal measures in the type area, the rose-
wood–Walloon coalfield, SE Queensland. Australian Coal Geology. 9, p. 16.
Free, D.L., 1989. Geology of the Main Range Volcanics and Hydrogeological Implications,
Toowoomba District, South East Queensland Toowoomba: The author .
Giambastiani, B.M.S., McCallum, A.M., Andersen, M.S., Kelly, B.F.J., Acworth, R.I., 2012. Un-
derstanding groundwater processes by representing aquifer heterogeneity in the
Maules Creek Catchment, Namoi Valley (New South Wales, Australia). Compréhen-
sion des processus d'écoulement souterrain par représentation de l'hétérogénéité
de l'aquifère du captage Maules Creek, vallée Namoi (Nouvelle-Galles du Sud,
Australie). 20, pp. 1027–1044.
Gómez, J.B., Gimeno, M.J., Auqué, L.F., Acero, P., 2014. Characterisation and model-
ling of mixing processes in groundwaters of a potential geological repository
for nuclear wastes in crystalline rocks of Sweden. Sci. Total Environ. 468–469,
791–803.
Hansen, A., 1999. Groundwater vulnerability and availability mapping of the Upper
Condamine River Catchment. Dept. of Natural Resources, Brisbane, Qld.: Queensland.
Hocking, M., Kelly, B.F.J., 2016. Groundwater recharge and time lag measurement through
Vertosols using impulse response functions. J. Hydrol. 535, 22–35.
Hooper, R.P., 2003. Diagnostic tools for mixing models of stream water chemistry. Water
Resour. Res. 39, 1055.
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, 344–351.
Huxley, W.J., 1982. The Hydrogeology, Hydrology and Hydrochemistry of the Condamine
River Valley Alluvium. Dept. of Applied Geology. Master. Queensland Institute of
Technology, Brisbane, p. 286.
Huxley, W.J., QWRC, 1982. An Assessment of the Groundwater Resources of the Tributary
Streams of the Upper Condamine River Catchment. Queensland Water Resources
Commission, Brisbane.
James, A.L., Roulet, N.T., 2006. Investigating the applicability of end-member mixing anal-
ysis (EMMA) across scale: a study of eight small, nested catchments in a temperate
forested watershed. Water Resour. Res. 42, W08434.
Jell PA, Queensland GSo. Geology of Queensland: Geological Survey of Queensland, 2013.
Joreskog, K.G., Klovan, J.E., Reyment, R.A., 1976. Geological Factor Analysis. Elsevier Scien-
tific Pub. Co.
Jurado, A., Mastroianni, N., Vàzquez-Suñé, E., Carrera, J., Tubau, I., Pujades, E., et al., 2012.
Drugs of abuse in urban groundwater. A case study: Barcelona. Sci. Total Environ. 424,
280–288.
Jurado, A., Vàzquez-Suñé, E., Soler, A., Tubau, I., Carrera, J., Pujades, E., et al., 2013. Appli-
cation of multi-isotope data (O, D, C and S) to quantify redox processes in urban
groundwater. Appl. Geochem. 34, 114–125.
Jurado, A., Vázquez-Suñé, E., Carrera, J., Tubau, I., Pujades, E., 2015. Quantifying chemical
reactions by using mixing analysis. Sci. Total Environ. 502, 448–456.
Kaiser, H.F., 1958. The varimax criterion for analytic rotation in factor analysis.
Psychometrika 23, 187–200.
Kalbus, E., Reinstorf, F., Schirmer, M., 2006. Measuring methods for groundwater and sur-
face water interactions: a review. Hydrol. Earth Syst. Sci. 10, 873–887.
Kelly, B.F.J., Timms, W., Ralph, T.J., Giambastiani, B.M.S., Comunian, A., McCallum, A.M., et
al., 2014. A reassessment of the lower Namoi catchment aquifer architecture and hy-
draulic connectivity with reference to climate drivers. Australian Journal of Earth Sci-
ences]–>Aust. J. Earth Sci. 1-11.
King, A.C., Raiber, M., Cox, M.E., 2014. Multivariate statistical analysis of hydrochemical
data to assess alluvial aquifer–stream connectivity during drought and flood:
Cressbrook Creek, southeast Queensland, Australia. Hydrogeol. J. 1-20.
Klohn Crippen Berger, 2014. Murray Darling Basin Authority, Upper Condamine Alluvium
Tributaries - Hydrogeological Assessment Stage II: Hydrogeological Conceptualisa-
tion. p. 452.
17
Lane, W.B., 1979. Groundwater Resources of the Condamine River Alluvium - An assess-
ment, Recharge and Groundwater Model Study. Civil Engineering Department. Mas-
ter of Engineering Science, University of Queensland, Brisbane, QLD.
Lewis, S., English, P., Sandow, J., Coram, J., 2008. In: Adelaide, S.A. (Ed.), Unravelling the
Mysteries of Groundwater Systems for the Australian Public - Improved
Hydrogeological Visualisation and Understanding with Computer-generated 3D
Models. International Conference on Water Resources and Environment Research,
4th Engineers Australia; Causal Productions, Modbury, SA, pp. 227–238 2008.
Liu, Y., Yamanaka, T., 2012. Tracing groundwater recharge sources in a mountain–plain
transitional area using stable isotopes and hydrochemistry. J. Hydrol. 464–465,
116–126.
Martinez, J.L., Raiber, M., Cox, M.E., 2015. Assessment of groundwater–surface water in-
teraction using long-term hydrochemical data and isotope hydrology: headwaters
of the Condamine River, Southeast Queensland, Australia. Sci. Total Environ. 536,
499–516.
Mikita, M., Yamanaka, T., Lorphensri, O., 2011. Anthropogenic changes in a confined
groundwater flow system in the Bangkok basin, Thailand, part I: was groundwater-
recharge enhanced? Hydrol. Process. 25, 2726–2733.
Miller, B., Lerch, R., Groves, C., Polk, J., 2015. Recharge mixing in a complex distributary
spring system in the Missouri Ozarks, USA. Hydrogeol. J. 1-15.
Monjerezi, M., Vogt, R.D., Aagaard, P., Saka, J.D.K., 2011. Hydro-geochemical processes in
an area with saline groundwater in lower Shire River valley, Malawi: an integrated
application of hierarchical cluster and principal component analyses. Appl. Geochem.
26, 1399–1413.
Morales-Casique, E., 2012. Mixing of groundwaters with uncertain end-members: case
study in the Tepalcingo-Axochiapan aquifer, Mexico. Hydrogeol. J. 20, 605–613.
Moya, C.E., Raiber, M., Cox, M.E., 2014. Three-dimensional geological modelling of the gal-
ilee and central Eromanga basins, Australia: new insights into aquifer/aquitard geom-
etry and potential influence of faults on inter-connectivity. Journal of Hydrology:
Regional Studies 2, 119–139.
Nakaya, S., Uesugi, K., Motodate, Y., Ohmiya, I., Komiya, H., Masuda, H., et al., 2007. Spatial
separation of groundwater flow paths from a multi-flow system by a simple mixing
model using stable isotopes of oxygen and hydrogen as natural tracers. Water Resour.
Res. 43, W09404.
Office of Groundwater Impact Assessment, 2016. Underground Water Impact Report for
the Surat Cumulative Management Area.
Owen, D.D.R., Cox, M.E., 2015. Hydrochemical evolution within a large alluvial groundwa-
ter resource overlying a shallow coal seam gas reservoir. Sci. Total Environ. 523,
233–252.
Oyarzún, R., Barrera, F., Salazar, P., Maturana, H., Oyarzún, J., Aguirre, E., et al., 2014. Multi-
method assessment of connectivity between surface water and shallow groundwa-
ter: the case of Limarí River basin, north-central Chile. Hydrogeol. J. 1-17.
Parkhurst, D.L., Appelo, C., 2013. Description of input and examples for PHREEQC version
3: a computer program for speciation, batch-reaction, one-dimensional transport,
and inverse geochemical calculations. US Geological Survey, p. 497.
Parkhurst, D.L., Charlton, S.R., 2008. NetpathXL, an Excel Interface to the Program
NETPATH. US Geological Survey, US Department of the Interior.
Plummer, L.N., Prestemon, E.C., Parkhurst, D.L., 1994. An interactive code (NETPATH) for
modeling NET geochemical reactions along a flow PATH, version 2.0. Water-
Resources Investigations Report.
QDME, 2008. Queensland Geological Mapping (polygonised vector) Data Regional & 1:
100 000 Sheet areas.
Raiber, M., White, P.A., Daughney, C.J., Tschritter, C., Davidson, P., Bainbridge, S.E., 2012.
Three-dimensional geological modelling and multivariate statistical analysis of
water chemistry data to analyse and visualise aquifer structure and groundwater
composition in the Wairau Plain, Marlborough District, New Zealand. J. Hydrol.
436–437, 13–34.
Raiber, M., Webb, J.A., Cendón, D.I., White, P.A., Jacobsen, G.E., 2015. Environmental iso-
topes meet 3D geological modelling: Conceptualising recharge and structurally-
controlled aquifer connectivity in the basalt plains of south-western Victoria,
Australia. J. Hydrol. 527, 262–280.
Ramos-Leal, J.A., Martínez-Ruiz, V.J., Rangel-Mendez, J.R., Alfaro de la Torre, M.C., 2007.
Hydrogeological and mixing process of waters in aquifers in arid regions: a case
study in San Luis Potosi Valley, Mexico. Environ. Geol. 53, 325–337.
Rezaei, M., Sanz, E., Raeisi, E., Ayora, C., Vázquez-Suñé, E., Carrera, J., 2005. Reactive trans-
port modeling of calcite dissolution in the fresh-salt water mixing zone. J. Hydrol.
311, 282–298.
Robins, N., Rutter, H., Dumpleton, S., Peach, D., 2005. The role of 3D visualisation as an an-
alytical tool preparatory to numerical modelling. J. Hydrol. 301, 287–295.
Rueedi, J., Purtschert, R., Beyerle, U., Alberich, C., Kipfer, R., 2005. Estimating groundwater
mixing ratios and their uncertainties using a statistical multi parameter approach.
J. Hydrol. 305, 1–14.
Smerdon, B., Ransley, T., 2012. Water Resource Assessment for the Surat Region. Canberra,
Commonwealth of Australia.
Stein, M.L., 1999. Interpolation of Spatial Data: Some Theory for Kriging. Springer Science
& Business Media.
Sun, L., Gui, H., 2015. Hydro-chemical evolution of groundwater and mixing between
aquifers: a statistical approach based on major ions. Appl Water Sci 5, 97–104.
Tan, P.-L., Baldwin, C., White, I., Burry, K., 2012. Water planning in the Condamine Alluvi-
um, Queensland: sharing information and eliciting views in a context of
overallocation. J. Hydrol. 474, 38–46.
Tonkin, M.J., Larson, S.P., 2002. Kriging water levels with a regional-linear and point-log-
arithmic drift. Ground Water 40, 185–193.
Tubau, I., Vàzquez-Suñé, E., Jurado, A., Carrera, J., 2014. Using EMMA and MIX analysis to
assess mixing ratios and to identify hydrochemical reactions in groundwater. Sci.
Total Environ. 470–471, 1120–1131.
18 J.L. Martinez et al. / Science of the Total Environment 574 (2017) 1–18
van Dijk, A.I.J.M., Beck, H.E., Crosbie, R.S., de Jeu, R.A.M., Liu, Y.Y., Podger, G.M., et al., 2013.
The millennium drought in Southeast Australia (2001–2009): natural and human
causes and implications for water resources, ecosystems, economy, and society.
Water Resour. Res. 49, 1040–1057.
Vanderzalm, J.L., Jeuken, B.M., Wischusen, J.D.H., Pavelic, P., Le Gal La Salle, C., Knapton, A.,
et al., 2011. Recharge sources and hydrogeochemical evolution of groundwater in al-
luvial basins in arid central Australia. J. Hydrol. 397, 71–82.
Vázquez-Suñé, E., Carrera, J., Tubau, I., Sánchez-Vila, X., Soler, A., 2010. An approach to
identify urban groundwater recharge. Hydrol. Earth Syst. Sci. 14, 2085–2097.
Voeckler, H.M., 2012. Modeling Deep Groundwater Flow through Fractured Bedrock in a
Mountainous Headwater Catchment Using a Coupled Surface Water-Groundwater
Model. Okanagan Basin, British Columbia.
Willey, E.C., 2003. Urban geology of the Toowoomba conurbation, SE Queensland,
Australia. Quat. Int. 103, 57–74.
Wilson, J.L., Guan, H., 2004. Mountain-block hydrology and mountain-front recharge.
groundwater recharge in a desert environment: the southwestern United States.
American Geophysical Union]–>Am. Geophys. Union 113–137.