Evolution of Dissolved Inorganic Carbon in Groundwater Recharged by Cyclones and Groundwater Age Estimations Using the 14C Statistical Approach

Available online at www.sciencedirect.com ScienceDirect

Geochimica et Cosmochimica Acta 220 (2018) 483–498 www.elsevier.com/locate/gca

Evolution of dissolved inorganic carbon in groundwater recharged by cyclones and groundwater age estimations using the 14C statistical approach

K.T. Meredith a,⇑, L.F. Han b, D.I. Cendo´n a, J. Crawford a, S. Hankin a, M. Peterson a, S.E. Hollins a

a The Environment, Australian Nuclear Science and Technology Organisation, New Illawarra Road, Lucas Heights, NSW 2234, Australia b Hydrology and Water Resources Department, Nanjing Hydraulic Research Institute, Guangzhou Road 223, P.O. Box 210029, Nanjing, China

Received 24 November 2016; accepted in revised form 2 September 2017; available online 12 September 2017

Abstract

The Canning Basin is the largest sedimentary basin in Western Australia and is located in one of the most cyclone prone regions of Australia. Despite its importance as a future resource, limited groundwater data is available for the Basin. The main aims of this paper are to provide a detailed understanding of the source of groundwater recharge, the chemical evolution 14 of dissolved inorganic carbon (DIC) and provide groundwater age estimations using radiocarbon ( CDIC). To do this we combine hydrochemical and isotopic techniques to investigate the type of precipitation that recharge the aquifer and identify 14 13 the carbon processes influencing CDIC, d CDIC, and [DIC]. This enables us to select an appropriate model for calculating radiocarbon ages in groundwater. The aquifer was found to be recharged by precipitation originating from tropical cyclones imparting lower average d2H and d18O values in groundwater (56.9‰ and 7.87‰, respectively). Water recharges the soil zone rapidly after these events and the groundwater undergoes silicate mineral weathering and clay mineral transformation processes. It was also found that partial carbonate dissolution processes occur within the saturated zone under closed system conditions. Additionally, the processes could be lumped into a pseudo-first-order process and the age could be estimated using 14 14 14 14 the C statistical approach. In the single-sample-based C models, C0 is the initial CDIC value used in the decay equation that considers only 14C decay rate. A major advantage of using the statistical approach is that both 14C decay and geochemical 14 14 processes that cause the decrease in CDIC are accounted for in the calculation. The CDIC values of groundwater were found to increase from 89 pmc in the south east to around 16 pmc along the groundwater flow path towards the coast indicating ages ranging from modern to 5.3 ka. A test of the sensitivity of this method showed that a 15% error could be found for the oldest water. This error was low when compared to single-sample-based models. This study not only provides the first groundwater age estimations for the Canning Basin but is the first groundwater dating study to test the sensitivity of the statistical approach and provide meaningful error calculations for groundwater dating. Crown Copyright Ó 2017 Published by Elsevier Ltd. All rights reserved.

Keywords: Carbon isotopes; Radiocarbon; Hydrochemistry; Canning Basin; Episodic recharge; Wallal Sandstone

1. INTRODUCTION

Groundwater is often the only reliable source of water in arid regions throughout the world. The Canning Basin ⇑ Corresponding author. located in Western Australia, is remote and one of the E-mail address: [email protected] (K.T. Meredith). https://doi.org/10.1016/j.gca.2017.09.011 0016-7037/Crown Copyright Ó 2017 Published by Elsevier Ltd. All rights reserved. 484 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 hottest and driest areas of Australia, experiencing extreme niques that measure the water sample directly (Fontes, temperatures (up to 49 °C) but it is also one of the most 1980) may prove more robust, and will provide a greater economically important. The prediction that average tem- understanding of the hydrological response of large peratures will rise by 1.5 °C by 2030 (McFarlane, 2015) will groundwater basins. directly increase the evaporation rates that greatly exceed Estimating groundwater age using radio-isotope tracers the often seasonal and highly variable rainfall and thus such as radiocarbon (14C) is important for any groundwater reduce the surface water availability for the region. resource assessment because they can offer guidance on the Although the direct impact of these changes on surface sustainability of a groundwater resource and can also be water resources is well known, the impact of longer drought used to calibrate groundwater flow models that can be used periods and higher intensity rainfall on groundwater stor- as tools to give an understanding of groundwater resources. age remains largely unknown. It has been identified that cli- But in order to use the 14C content of dissolved inorganic 14 mate extremes will influence groundwater recharge (Taylor carbon ( CDIC) to calculate groundwater ages, the initial 14 14 14 et al., 2013) and that there is a need to better understand CDIC value ( C0) must be known. The initial CDIC available groundwater storage as a result of rainfall intensi- value can be estimated provided that the carbon sources fication (Taylor, 2014). and reactions that affect carbon mass transfer are known The Canning Basin is the largest sedimentary basin in or there is strong geochemical evidence for the assumptions Western Australia and is second in size to the Great Arte- used (e.g. Ingerson and Pearson, 1964; Gonfiantini, 1972; sian Basin in Australia. Gravity Recovery and Climate Geyh, 2000; Han and Plummer, 2013; Plummer and Experiment (GRACE) satellite mission data showed that Glynn, 2013). These single-sample-based models rely on the Canning Basin is a variably stressed aquifer system with mass balances of major carbon species or carbon isotopes a positive value of recharge and negative value of use (14C and 13C) of DIC in a single water sample. Therefore, (Richey et al., 2015). The high stress ratio given to the Basin an understanding of evolution of DIC in groundwater is suggests that about 150% more water is being depleted than important for the successful application of 14C in estimating is naturally available. This assessment contradicts other groundwater age. The graphical analysis method Han et al. global studies that suggest that the aquifers of the Canning (2012) can be used to aid in revealing the complexity of the Basin are renewable groundwater resources (Gleeson et al., geochemical environment, conceptualising the processes 14 2012). that affect CDIC in aquifers, and, determining which The conflicting global scale studies highlight our limited approach is most appropriate for the estimation of the ini- 14 knowledge of the hydrological balance of the Canning tial CDIC. Basin. Rainfall sources to the Canning Basin are well The main aims of this paper are to provide a detailed known, particularly in the coastal areas, as being influenced understanding of the source of groundwater recharge, the by both tropical maritime air from the Indian Ocean and chemical evolution of DIC and provide the first groundwa- continental air from inland. These weather patterns result ter age estimations for the Canning Basin. To do this we in extreme rainfall conditions ranging from severe droughts combine hydrochemical and isotopic techniques to investi- to cyclonic rainfall events. The rainfall in summer can be gate the type of precipitation that recharge the aquifer 14 13 widespread, associated with cyclonic weather from tropical and identify the carbon processes influencing CDIC, d - low-pressure systems with centres formed to the north-west CDIC, and [DIC]. This enables us to select an appropriate of the coast (Sturman and Tapper, 1999; Dogramaci et al., model for radiocarbon dating of DIC in groundwater. In 2012). The Pilbara coast, where the Canning Basin is addition, this paper not only presents the first hydrochem- located, experiences more cyclones than any other part of ical and isotope dataset for groundwater in the Canning Australia. Since 1910, there have been 48 cyclones (i.e. on Basin but is also the first study to use the statistical mod- average about one every two years (BOM, 2016)). The win- elling approach for calculating a 14C age for groundwaters ter rainfall for the region is relatively low, with the majority that were recharged by cyclone sourced rainfall. Further- of which (about 70–90%) resulting from north-west cloud more, this is the first study to test the sensitivity of the sta- bands (Wright, 1997; Sturman and Tapper, 1999). Based tistical approach and provide meaningful error calculations on the climatic data, it seems reasonable to suggest that for this groundwater dating method. cyclone events are the dominant groundwater recharge sources but understanding how this water is recharged 2. ENVIRONMENTAL SETTING and the volume being recharged is still poorly understood for the region. 2.1. Study site Sustainable groundwater use requires quantification of fluxes and groundwater storage (Gleeson et al., 2016). A The West Canning Basin (WCB) covers an area of groundwater flow model was created for the site 10,000 km2 and is a subset of the Canning Basin, Western (Aquaterra, 2009), which can be used to give an indication Australia. The study site is situated between Pardoo Station of the groundwater storage volume. However, estimating in the north-western corner and Shay Gap in the south- groundwater storage is challenging in most arid zone sys- eastern corner (Fig. 1). The nearest high quality long-term tems (Scanlon, 2000) and models require data for valida- climate monitoring site is Port Hedland (Bureau of Meteo- tion. Monitoring and sampling groundwater recharge rology site number: 004032) where a mean annual rainfall from high volume rainfall events is difficult especially in of 318.5 mm yr1 (1942–2013) and potential evaporation remote arid regions. This is where isotope hydrology tech- of 3285 mm yr1 (1967–2013) was recorded, this is repre- K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 485

Fig. 1. (a) Elevation and location map of the study site with regards to the location of the West Canning Basin containing the hydrology, well locations, meteorological stations, and the extent of the unconfined and artesian areas as identified by Aquaterra (2010) for the Wallal Sandstone. The elevation ranges from 0 m asl depicted in blue along the coast line to 275 m asl in brown for inland areas in the vicinity of Shay Gap and (b) the south (A’) to north (A) geological cross section after Haig (2009) showing the location of wells and the screen depths shown as light grey rectangles. sentative for the study area including the recharge area. The The terrigenous sediments are dominated by red siliclastic lowest mean monthly minimum temperature of 12.3 °C and lithoclastic sands and clay mineral muds, derived from occurs in July (1948–2013) and the highest mean monthly the Precambrian hinterland (Semeniuk, 1996). Soil profiles maximum temperature of 36.7 °C, occurs in March are weakly developed because of limited vegetation cover. (BOM, 2013). The area has a mean daily relative humidity The vegetation present includes Mulga (Acacia aneura), of 41%. woodlands and shrublands with hummock grass on the The study site is situated on a plain ranging in elevation coast and inland over the sandplain (McFarlane, 2015). from sea level at the coast to 100–200 m above sea level inland (Fig. 1). The coastal plain is a grassed area with tidal 2.2. Hydrogeology salty flats and recent sand dunes, while the inland area is covered by terrigenous and pedogenic carbonate sediments The West Canning Basin (WCB) is a multilayered aqui- with occasional mesas and fixed seif dunes (Leech, 1979). fer system, with the main aquifers being the Broome and 486 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498

Wallal Sandstones. These units are separated by the Jar- Groundwater discharge is observed near the coast as small lemai Siltstone, which acts as an aquitard (Fig. 1b). The springs but predominately occurs as submarine discharge Wallal Sandstone is the focus of this study. It is present and ranges between 4.6 107 and 2.4 108 m3 yr1 across the entire Basin as a semi-confined artesian aquifer (McFarlane, 2015). that comprises sandstone with rare siltstones and gravel interbeds. Sediments are very coarse to fine grained, poorly 3. METHODS consolidated, and are fawn to light-grey in colour. It comprises poorly to well sorted, sub-angular to rounded grains. Four field trips in December 2008, October 2009, June The quartz sandstone contains traces of carbonate miner- 2012 and May 2013 were conducted to collect groundwater als, pyrite, black heavy minerals and rose coloured garnets samples. During these, 28 groundwater samples were col- and forms the basal unit of a marine transgression (Leech, lected from groundwater monitoring wells for hydrochemi- 1979). The Wallal Sandstone unconformably overlies Pre- cal analysis. Sampling methods differed depending on cambrian rocks and reaches a maximum recorded thickness whether the aquifer was pressurised, and if not, the method of 420 m to the north-east of the project area (not shown in depended on the depth of the standing water levels (SWLs). cross section). It is confined by the Jarlemai Siltstone, In the case of artesian wells, groundwater was purged except in the south where it is in direct contact with the through the main pressure valves. A minor outflow valve Broome Sandstone (depicted in the southern section of was plumbed into a flow cell allowing for simultaneous the cross section in Fig. 1b). The aquifer is artesian along purging and monitoring of field parameters. Wells with the coastal zone, with potentiometric heads of up to 50 m SWLs (<44 m bgs) were sampled using low-flow methods above the ground surface. Groundwater flow direction fol- (Puls and Barcelona, 1996) and others were sampled using lows a general west-north-west direction (Fig. 2) a Grundfos MP1 submersible pump following the standard (Aquaterra, 2009), which is oblique to the available sam- 3 well volume method. pling transects. Groundwater samples were generally collected via an in- 13 Recharge to the Wallal Sandstone can only occur inland line, 0.45 lm filter, with d CDIC samples further filtered beyond its artesian boundary and most likely where the through 0.22 lm. Total alkalinity was determined in the confining layer of the Jarlemai Siltstone is absent near the field by acid-base titration using a HACH digital titrator. limit boundary of the Basin to the south in Fig. 1. The dom- Samples were collected, preserved and measured according inant lateral groundwater contribution from the wider Can- to Appendix A. Specifically, cations and anions were mea- ning Basin recharge area is not investigated in this study but sured by inductively coupled plasma – atomic emission is likely to originate to the south-east of the study site as spectroscopy and ion chromatography, respectively. The suggested by the potentiometric head configuration. d18O and d2H were analysed by IRMS and are reported

Fig. 2. The spatial distribution of Cl (mmol L1) in groundwaters for the study area compared with 2 m piezometric surface contours for the Wallal Sandstone (Aquaterra, 2009). Polygons represent groupings identified by the cluster analysis. Note: group 4a inland includes all the wells that are not in a polygon. K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 487 as per mil (‰) deviations from the international standard base in the PHREEQC programme (Parkhurst and Appelo, V-SMOW and were reproducible to ±0.1‰ and ±1.0‰, 1999). We use the graphical method developed by Han et al. respectively. (2012) to interpret groundwater evolution based on the con- The d13C signatures of waters were analysed by IRMS centration and the isotopic composition of the DIC. The and results were reported as ‰ deviation from the interna- 87/86Sr ratios were converted into per mil (d87Sr) tional carbonate standard, NBS19 with a precision of (Elderfield, 1986). All the methods including references, col- ±0.1‰ according to methods reported in Meredith et al. lection vessels, reference material and errors are reported in 14 (2016). The CDIC content was determined by accelerator Appendix A. mass spectrometry after samples were processed according to the methods outlined in Meredith et al. (2012). Briefly, 4. RESULTS the total DIC or carbonate was processed into CO2 by acid- ifying the samples and extracting the liberated CO2 gas. The A range of hydrochemical (major ions and trace ele- CO2 sample was then heated with CuO, Ag and Cu wire, at ments) and environmental isotopes were measured on 600 C for 2 h and then converted into graphite by reducing groundwaters from the Wallal Sandstone over the duration it with excess hydrogen gas in the presence of an iron cata- of the study. Four major groups of groundwaters were lyst at 600 C. The 14C results reported from the laboratory identified by using classical hydrochemical X-Y plot rela- were in percent Modern Carbon (pMC) normalised against tionships that show the addition or removal of groundwater the d13C of the graphite. We use the ‘un-normalised’ 14C ions and can then be used to infer geochemical processes values for groundwater and report them as pmc which that have led to the evolution of groundwater. The location was calculated using the 14C/12C ratio (Plummer and of the different hydrochemical groups were similar to those Glynn, 2013). identified with cluster analysis. The relationship between Cations and anions were assessed for accuracy by eval- these ions also relates to the geographical location of the uating the charge balance error percentage (CBE%; groundwaters in the aquifer therefore the cluster analysis Table 1). Most samples (except WCB15Y = 7.8%) fell was used as an additional line of evidence for hydrochemi- within the acceptable ±5% range. Hierarchical cluster anal- cal interpretation. These groups include: (1) Group 1 - the ysis was carried out to validate the hydrochemical observa- well located in the far north-west (WCB04), (2) Group 2 - tions and the following parameters were selected; HCO3, wells located in the north-west (WCB08, WCB09, Cl, SO4, SiO2, Na, Ca, Mg, K and Sr. To reduce the impact WCB10), (3) Group 3 - wells located inland (PB1 and of the magnitude of each parameter, the mean and standard PB3) and (4) Group 4 - wells also located inland and along deviations of each parameter were calculated, then each the coast (Fig. 2). Based on the cluster analysis, Group 4 parameter’s value was normalised by subtracting the mean was further sub-divided into group 4a located inland and dividing by the standard deviation. Following this, (WCB20Y, WCB18, WCB19, WCB15Y, MMCA03E, hierarchical cluster analysis was carried out, where the sim- WCB23B), group 4b (WCB25X, WCB25Y, Cooragoora, ilarity between two sites was calculated using Euclidian dis- WCB24Y, WCB21Y) and group 4c located on the coast tance between the parameters, with each parameter having (WCB22, WCB17 and Pardoo) (Fig. 2). Groundwaters the same weighting. In hierarchical cluster analysis, initially were also assessed with distance from Shay Gap (Fig. 1). each object is assigned to its own cluster and the distances This was done because the major recharge area appears between clusters are calculated. The algorithm then pro- to be located to the south east of the study site according ceeds iteratively; at each stage joining the two clusters with to the potentiometric contours (Fig. 2). This transect does the shortest distance. The distances between the resulting not strictly represent the groundwater flow path but a flow clusters are then re-calculated and the iterations are contin- path from WCB23B ? WCB20 ? Pardoo wells is shown in ued until all clusters have been merged into one cluster. The Figs. 4–6 for reference. optimum number of clusters can be chosen when the cost of The groundwater Cl concentration in the Wallal Sand- merging two clusters is high (e.g. when objects with large stone is generally below 5 mmol L1 but rises to over difference are grouped into the one cluster) and/or by anal- 10 mmol L1, reaching a high of 21.4 mmol L1 in the ysis the cluster dendrogram (cluster tree). The cluster anal- western coastal area (Groups 1 and 2) (Table 1, Fig. 2). 14 2 18 ysis was run with and without CDIC, d H and d Oas Groundwaters from Group 3, 4a and 4b have the lowest variables, resulting in no change in the groupings. A princi- Cl concentrations below 3.5 mmol L1 (Fig. 2) and the low- ple component analysis (PCA) was completed on ground- est ion concentrations (Table 1). The Na/Cl ratios are gen- waters from cluster groups 3, 4a and 4b (Fig. 2). erally around 1 for Group 1 and 2 groundwaters but Parameters used include HCO3, Cl, SO4, SiO2, Na, Ca, increase to greater than 1 in Groups 3 and 4, indicating Mg, K and Sr. Principle Component Analysis (Jollife, there is a source of Na relative to Cl in these groundwaters 1986) and cluster analysis (using the K-means clustering (Fig. 3a). Groundwaters located closer to the coast (Group algorithm; (Hartigan and Wong, 1979) were performed 4c) have higher Cl concentrations between 3.5 and using the R programming environment (Venables et al., 8 mmol L1. The Cl/Br ratios for all groundwaters (except 2016). those from Groups 1 and 2) suggest Cl is derived from mar- Saturation indices (calcite, dolomite and siderite), con- ine aerosols with ratios close to marine signatures of 655 centration of dissolved inorganic carbon [DIC], carbon (Fig. 3a). The increase in Cl concentration across the site 2 dioxide [CO2], carbonate [CO3 ] and bicarbonate [HCO3 ] is broadly reflected in the concentration of other ions, with were calculated using the WATEQ4F thermodynamic data- an excellent correlation (r2 greater than 0.9) with Cl, Na, 8 ..Mrdt ta./Gohmc tCsohmc ca20(08 483–498 (2018) 220 Acta Cosmochimica et Geochimica / al. et Meredith K.T. 488

Table 1 Hydrochemical results for groundwater samples.

ID Date Depth Distance T pH DO Na Ca Mg K Sr Cl HCO3 SO4 SiO2 CBE m bgs km mg L 1 mmol L 1 mmol L 1 mmol L 1 mmol L 1 umol L 1 mmol L 1 mmol L 1 mmol L 1 mmol L 1 % Cooragoora 7/06/2012 240 39 34.4 6.6 0.05 4.2 0.4 0.4 0.2 2.4 2.4 2.5 0.5 0.1 1.1 MMCA03E 25/05/2013 102 5 30.72 6.4 5.19 2.3 0.6 0.6 0.1 2.3 2.5 1.8 0.1 0.5 4.1 Pardoo 7/06/2012 142 35 34.6 6.5 0.05 6.7 0.8 0.7 0.2 4.0 5.4 2.5 1.0 0.1 0.1 PB1 12/02/2008 112 2 33.5 6.5 6.71 3.0 0.5 0.6 0.1 3.3 3.1 1.6 0.2 1.0 2.9 PB3 12/02/2008 111 2 33.6 6.2 6.54 1.7 0.4 0.4 0.1 2.4 1.7 1.0 <0.1 1.0 7.7 PB3 10/12/2009 111 2 33.4 6.4 6.23 1.7 0.4 0.4 0.1 2.3 1.7 1.0 <0.1 1.0 8.8 WCB04 8/06/2012 82 10 32.3 6.3 0.4 18.7 2.1 1.8 0.5 14.0 20.6 1.1 2.5 0.1 0.6 WCB04 12/04/2008 82 10 32.4 6.2 1.18 18.7 2.1 1.9 0.5 13.7 21.4 0.9 2.4 0.2 0.0 WCB08 12/03/2008 85 26 32.8 6.3 0.68 10.0 1.5 1.3 0.3 7.8 10.7 2.5 1.7 0.3 -2.2 WCB09 12/03/2008 136 37 32.5 6.2 0.21 9.1 1.6 1.3 0.3 8.2 12.7 0.9 0.8 0.2 0.1 WCB10 12/04/2008 72 18 31.9 6.3 0.34 11.7 1.7 1.4 0.3 9.2 14.1 1.2 1.5 0.3 -0.2 WCB15Y 5/06/2012 107 16 34.7 6.8 4.21 1.9 0.7 0.5 0.1 2.6 1.7 1.9 0.1 0.5 7.8 WCB17 14/10/2009 140 39 34.28 6.8 0.24 6.1 0.5 0.5 0.2 2.7 4.2 2.8 0.8 0.3 -3.2 WCB17 12/03/2008 140 39 34.2 6.5 0.43 6.5 0.5 0.5 0.2 2.7 3.9 4.0 0.8 0.3 -5.3 WCB18 23/05/2013 119 3 31.98 6.5 3.98 3.0 0.8 0.6 0.1 3.2 3.1 2.3 0.1 0.6 1.4 WCB19 23/05/2013 150 15 32.54 6.6 2.55 1.8 0.7 0.5 0.1 2.2 1.8 2.2 0.1 0.5 2.8 WCB20Y 24/05/2013 195 23 32.34 6.8 3.79 2.8 0.6 0.4 0.1 1.9 1.9 2.5 0.1 0.4 2.7 WCB21Y 22/05/2013 217 30 35.21 6.4 2.51 3.2 0.4 0.3 0.1 2.0 2.1 2.0 0.3 0.3 1.9 WCB22 12/03/2008 144 37 32.9 6.4 0.77 7.0 0.7 0.6 0.2 4.1 4.8 3.4 1.0 0.3 -1.3 WCB23B 25/05/2013 106 18 33.62 6.4 0.06 2.0 0.7 0.4 0.1 2.2 1.5 2.1 0.2 0.4 2.5 WCB24Y 21/05/2013 204 28 34.77 6.5 0.6 2.7 0.4 0.2 0.1 1.6 1.5 2.0 0.2 0.2 0.8 WCB25X 6/06/2012 262 38 35.12 7.0 0.03 3.8 0.3 0.2 0.1 1.5 1.7 2.5 0.3 0.1 0.9 WCB25Y 6/06/2012 324 38 34.6 6.8 0.08 4.2 0.3 0.3 0.1 1.7 2.5 2.3 0.4 0.1 -0.1 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 489

Fig. 3. Bivariate plots of (a) Cl/Br vs. Na/Cl (molar concentration) and (b) d2H vs. d18O compared with the Darwin LMWL_PWLSR (d2H‰ = 7.69d18O + 9.75 [black solid line]), Local Meteoric Water Line (LMWL) for rainfall events >20 mm for Hamersley Basin (n = 8; d2H‰ = 7.0d18O + 4.8 [grey dashed line]) (Dogramaci et al., 2012), and a Local Evaporation Line (LEL) (d2H‰ = 5.2d18O – 14.4 [black dashed line] (Dogramaci et al., 2012).

Fig. 4. Distance from the south eastern corner of the study area (Shay Gap in Fig. 1) for groundwaters from cluster groups 3 and 4 with 2 2 2 respect to (a) SiO2 concentration (r = 0.8), (b) Ca and Mg concentration (both r = 0.0) and (c) Na concentration (r = 0.6). The linear fit is depicted by a solid black line. This figure does not represent the groundwater flow path as shown in Fig. 2. A groundwater flow path from wells located at WCB23B ? WCB20 ? Pardoo shows how parameters change.

SO4, Br, Ca, Mg, K, Sr and Rb. The high correlation coef- dard deviation of 1.7 and 0.30, respectively, suggesting a ficients imply a similar source of ions for the fresher waters. common origin of water. Groundwaters for the region plot The Cl/Br and Na/Cl ratios also show the difference in to the right of the Darwin Local Meteoric Line (LMWL) groundwaters from Groups 1 and 2 which have higher (Fig. 3b) on a regression line described by d2H = 4.08 Cl/Br ratios (800–1200) suggesting an alternative source d18O – 24.72. The only rainfall study for this region was of salinity for this section of the aquifer but the Na/Cl from Dogramaci et al. (2012) for rainfall events >20 mm ratios are within the marine range (0.8–1) (Fig. 3a). These (d2H‰ = 7.0d18O – 4.8), this line is similar to the Darwin finding are very important because they show that there is LMWL (Fig. 3b). not a single source of Cl within the aquifer. A principle component analysis was used in this study to The d2H and d18O values have an average of 56.9‰ confirm the relationships that were identified from classical and 7.87‰, respectively (n = 33) (Table 2) with a stan- hydrochemical graphical methods and were only under- 490 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498

Fig. 5. Distance from the south eastern corner of the study area (Shay Gap in Fig. 1) for groundwaters from cluster groups 3 and 4 with respect to (a) Na/Cl (r2=0.6), (b) d87Sr values (r2=0.7) and (c) Sr concentration (r2=0.0). The linear fit is depicted by a solid black line. This figure does not represent the groundwater flow path as shown in Fig. 2. A groundwater flow path from wells located at WCB23B ? WCB20 ? Pardoo shows how parameters change.

Fig. 6. Distance from the south eastern corner of the study area (Shay Gap in Fig. 1) for groundwaters from cluster groups 3 and 4 with 2 13 2 14 2 respect to (a) DIC (r = 0.3), (b) d CDIC (r = 0.4) and (c) CDIC (r = 0.9). The linear fit is depicted by a solid black line. This figure does not represent the groundwater flow path as shown in Fig. 2. A groundwater flow path from wells located at WCB23B ? WCB20 ? Pardoo shows how parameters change.

taken on groundwaters from cluster Groups 3, 4a and 4b. were negatively loaded, except SiO2. Cl had the highest This was done to help identify the major hydrochemical absolute loading with Ca, Mg and Sr being the most processes leading to the formation of the observed ground- related. water chemistry across the site. Component 1 (PCA1) The d87Sr values of groundwater from the Wallal Sand- explained 50% of the variation. Larger SiO2, Ca, Mg and stone are signiﬁcantly elevated (+6.6 to +13.7‰) compared Sr values were accompanied by lower HCO3,SO4,Na to seawater (0‰). Even though these values are high, they and K values. The positive loadings are associated with a are not to the levels found for the inland Pilbara region linear decrease in SiO2 concentrations, ranging from groundwaters (+27.2 to +40.2‰)(Dogramaci and 1 mmol L1 at PB1 and PB3, decreasing towards the coast Skrzypek, 2015). No relationship between Sr concentration to 0.1 mmol L1 (r2 = 0.8). The negative weighted variables and d87Sr values (R2 = 0.002) is observed in this study. 87 of Na, HCO3, K and SO4 increase in concentration similar However, an increase in Na/Cl ratios (Fig. 5a) and d Sr to Na concentration with distance (r2 = 0.6) (Fig. 4c). Com- values (Fig. 5b) is observed with distance from the Shay ponent 2 explained 28% of the variation and most variables Gap area (r2 = 0.6 and 0.7, respectively), suggesting a K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 491

Table 2 Environmental isotope results for groundwaters samples (Uncert. = 3H uncertainty, QL = 3H quantiﬁcation limit). 13 87 3 2 18 14 ID Date sampled d CDIC d Sr H Uncert. QL d H d O CDIC DIC SIcal ‰‰TU TU TU ‰‰pmc mmol L1 Cooragoora 7/06/2012 10.3 13.7 0.32 0.05 0.27 56.8 7.94 27.3 3.6 1.3 MMCA03E 25/05/2013 11.0 9.1 0.02 0.03 0.19 57.1 7.94 75.3 3.4 1.6 Pardoo 7/06/2012 9.4 12.9 53.7 7.40 16.0 3.9 1.2 Pardoo 2014 56.9 7.76 PB1 12/02/2008 13.1 60.95 8.35 77.3 2.7 1.5 PB1 10/12/2009 0.04 0.06 0.3 PB3 12/02/2008 15.5 61.18 8.2 81.7 2.3 2.1 PB3 10/12/2009 12.2 0.06 0.06 0.3 57.4 8.43 1.7 1.8 WCB04 8/06/2012 12.3 12.4 0.35 0.03 0.14 56.4 7.79 0.7 2.3 1.5 WCB04 12/04/2008 15.4 59.32 7.71 0.3 2.0 1.7 WCB04 2014 56.3 7.76 WCB08 12/03/2008 12.4 56 7.13 3.8 4.8 1.2 WCB08 2014 53.2 7.25 WCB09 12/03/2008 17.8 58.33 7.95 0.5 1.8 1.7 WCB09 2014 57.3 7.89 WCB10 12/04/2008 14.6 55.25 7.59 0.5 2.3 1.5 WCB10 2014 55.6 7.47 WCB15Y 5/06/2012 12.2 7.3 0.65 0.05 0.16 58.6 8.32 69.7 2.5 1.0 WCB17 12/03/2008 10.6 56.15 7.65 22.5 6.2 1.2 WCB17 14/10/2009 9.9 0.06 0.06 0.3 55 7.77 3.6 1.0 WCB17 2014 56.1 7.95 WCB18 23/05/2013 10.0 7.2 0.03 0.03 0.19 57.0 8.15 89.0 3.7 1.2 WCB19 23/05/2013 12.9 7.1 0.01 0.03 0.19 57.8 8.19 61.6 3.4 1.2 WCB20Y 24/05/2013 11.0 8.4 0.01 0.03 0.19 58.4 7.98 54.8 3.2 0.9 WCB21Y 22/05/2013 10.5 9.6 0.07 0.02 0.15 57.7 7.98 45.0 3.4 1.5 WCB21Y 2014 56.9 8.09 WCB22 12/03/2008 10.8 54.81 7.58 16.8 5.9 1.2 WCB22 2014 54.2 7.54 WCB23B 25/05/2013 12.3 6.7 0.02 0.03 0.19 57.8 8.02 49.6 3.8 1.4 WCB24Y 21/05/2013 12.4 9.2 0.01 0.03 0.15 57.2 8.02 44.6 3.2 1.5 WCB24Y 2014 57.0 8.07 WCB25X 6/06/2012 11.1 11.8 0.01 0.02 0.14 56.7 8.08 32.2 3.0 1.0 WCB25Y 6/06/2012 10.9 13.5 0.04 0.04 0.14 56.2 7.95 29.9 3.1 1.3 WCB25Y 2014 56.6 7.96

source of Na together with the addition of radiogenic Sr. detectable 3H (i.e. above the quantification limit of 0.2 TU, The Sr concentrations are generally relatively consistent Table 2) was WCB15 (0.7 TU). Except for this sample as shown with the trend line in Fig. 5c, unlike the d87Sr which is located inland near the unconfined section of the values. Wallal Sandstone, these results suggest groundwater has The d87Sr values suggest that the weathering of old not experienced any recent recharge (i.e. in contact with rubidium-rich primary silicate minerals (Faure and the atmosphere in the past approximately 50 years). Mensing, 2005) is a dominant process in this system. The addition of radiogenic Sr into the groundwater occurs close 5. DISCUSSION to the unconfined area of the Wallal Sandstone in the south east corner (near Shay Gap) and increases towards the 5.1. Groundwater recharge source coast (Fig. 5b). However we do not see an increasing trend in SiO2 or Sr concentrations (Figs. 4a and 5c) as would be Cyclonic rainfall events are likely to be the dominant expected if silicate weathering is occurring. This trend does source of groundwater recharge to the system, which can not initially support silicate mineral weathering processes. be identified in the rainfall records for the region but also 2 18 However, the loss of SiO2 from aqueous phase commonly from the low d H and d O values in groundwater (average occurs in clay mineral transformation reactions d18O and d2H values of 7.87‰ and 56.9‰ (n = 33), (Langmuir, 1997), which could account for the decrease respectively). The negative values of the groundwater can in SiO2 concentration (Fig. 4a), as we see in our data. be explained by studies such as Zwart et al. (2016) who 13 The [DIC] and d CDIC values of the groundwaters in undertook isotopic measurement of rainfall during two the Wallal Sandstone increase with distance from Shay monsoonal events in northern Australia. They observed 14 Gap (Fig. 6a and b) and conversely, the CDIC values that the isotopic composition of precipitation (range 129 decrease (Fig. 6c). The only groundwater sample to contain to 6‰ for d2H and 17 to +1‰ for d18O) was related 492 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 to the size and activity of the convective envelope, with the of the Wallal Sandstone occurs at the south eastern bound- most negative values occurring when eastward and west- ary (Figs. 1b and 2) providing potential pathways for ward moving precipitation systems merged over the mea- focussed groundwater recharge from cyclonic events. surement site. The isotopic composition of precipitation Understanding this recharge pathway is important for trac- from tropical cyclones, similar to those experienced across ing the evolution of DIC into the Wallal Sandstone. Fur- the study site, have not been studied for this area but are thermore, the Wallal Sandstone has been found to suggested to be typically more depleted in 18O and 2H than contain rare siltstones and gravel beds but also traces of any other precipitation (e.g. Lawrence et al., 2002; carbonate minerals and pyrite (Leech, 1979). Fudeyasu et al., 2008; Munksgaard et al., 2015). The presence of carbonate minerals suggests that car- Long-term isotopic data for precipitation was not avail- bonate dissolution has the potential to influence the hydro- able for this site therefore, to investigate the impact of trop- chemical evolution of groundwater DIC. Interestingly, ical cyclones, we use the Global Network of Isotopes in there is no firm evidence for this process in the saturation Precipitation (GNIP) data from Darwin (1500 km north- indices of carbonate minerals such as calcite and dolomite east from the study area). Even though this is a large dis- in the groundwater samples because they are under- tance from our study site it is used because it represents saturated with respect to these minerals. But when we look 14 13 the only Australian rainfall dataset to potentially contain at the relationship between CDIC, d CDIC, and [DIC] such precipitation events resulting from a cyclone. Using (Fig. 7), the various geochemical processes that may all the available data from Darwin, the precipitation influence the carbon chemistry of the groundwater system weighted average d18O and d2H values were calculated to (i.e. [DIC]), and the carbon isotopic composition of DIC 14 13 be 5.51‰ and 32.6‰, respectively. This value is much ( CDIC, d CDIC) can be identified. 13 higher than groundwaters from the study site and they also, Firstly, we see the changes in d CDIC and [DIC] caused plot to the right of the Darwin LMWL (Fig. 3b), suggesting by geochemical processes that involve the addition or they are not the source waters for groundwater for the removal of DIC with respect to enrichment or depletion region. But when we take a subset of data from Darwin of 13C(Fig. 7I). Fig. 7I does not show 14C decay, this is 13 for those months during which a cyclone was recorded, because it does not affect d CDIC and [DIC], therefore, the precipitation weighted average d18O and d2H values any changes in these parameters is the results of other phys- 14 were more depleted than the average value (6.42‰ and ical/chemical process(es) that may alter CDIC in the 40.9‰, respectively). However these values are still less absence of 14C decay (e.g. mixing of waters with different 13 depleted in heavy isotopes when compared to groundwaters d CDIC and [DIC] values or addition or removal of carbon from our study site suggesting these types of rainfall events caused by physical/chemical processes) (Han et al., 2012; do not represent recharge. The difference in these values can Han and Plummer, 2016). From Fig. 7I we see the general 13 be explained by the studies undertaken by Lawrence and trend is an increase in d CDIC with increasing [DIC]. Gedzelman (1996) and Gedzelman et al. (2003) that the iso- Here we suggest that the main reason for the increase of topic composition of precipitation (d18O and d2H) while DIC in groundwater is carbonate dissolution according to depleted relative to other rainfall was still varied and not the following (1): consistent within cyclone systems. 2þ CO ð Þ þ MeCO ð Þ þ H O ¼ 2HCO ð Þ þ Me ð1Þ Interestingly, when we add an Evaporation Line (EL) 2 aq 3 s 2 3 aq ðaqÞ derived from surface water sampling in a similar climatic where Me is generally Ca or Mg. The subscripts aq and s environment (Dogramaci et al., 2012)(Fig. 3b) we see a represent dissolved and solid states, respectively. The con- much closer relationship with groundwater samples. The centration of CO2(aq) and HCO3(aq) are equal at pH of 6.4 2 18 slight increase in d H and d O along the evaporation line in dilute aqueous solutions at 25 °C. Thus, the [DIC] may does suggest minor evaporation of the recharge water but increase if the CO2(aq) reacts further with carbonate miner- not open-water evaporation which would display much als (Eq. (1)). If soil CO2 in the recharge area were the only ‰ higher values (i.e. towards 0 ). Based on these results we source for CO2(aq) in Eq. (1), we would see an increase in suggest that groundwater recharges fairly quickly through pH value with increasing [DIC] and decreasing [CO2(aq)], 2 18 the unsaturated zone because the d H and d O cyclonic and from the material balance the maximum final concen- signal is retained in the groundwater, even under extreme tration of DIC would not exceed 1.5 times of the initial temperatures. Therefore, we suggest that the observed sig- [DIC] value. At maximum conversion of CO2(aq) (i.e. nal in the groundwater is produced by the mixing of evap- [DIC] approaches 1.5 times of its initial value), the resultant orated soil water contained in the unsaturated zone and the pH value would be greater than 8. But in fact, we do not see incoming recharge water before it reaches the water table. significant changes in pH with changing [DIC] (Table 1) Further unsaturated zone soil profile sampling would be and [DIC] has increased more than 1.5 times of the initial required to confirm this. value (Table 1 and Fig. 6a). Therefore, the increase in [DIC] is most likely to be caused by the reaction of carbon- 5.2. Evolution of DIC ate minerals with CO2(aq) that was added to the system during DIC evolution, and in addition to soil CO2. Groundwater flow direction follows a general west- To eliminate all the possible processes influencing DIC north-west direction (as shown with arrows in Fig. 2) and evolution, first we assume that the addition of CO2 has the dominant recharge area is to the south-east of the study resulted from the oxidation of fossil organic matter in the site. However, within the study area an unconfined section aquifer and carbonate dissolution (with a d13C value of K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 493

14 13 Fig. 7. The CDIC- d CDIC-[DIC] diagrams (Han and Plummer diagrams; Han and Plummer, 2013) of the data. See text for explanation of end members (A and S).

0‰). Under these assumptions, the added DIC should have Thus, we see that the DIC is the result of at least two 13 a d C value of approximately half of the soil CO2 (i.e. 0.5 geochemical processes. These could be carbonate dissolu- 13 d C(CO2)). The soil CO2 was calculated to be between tion caused by ion exchange on clay minerals and the addi- 13 18.4 and 16.9‰ for d C(CO2) (see Appendix B). There- tion of CO2 from oxidation of fossil organic matter in the 13 fore, the added DIC should have a d CDIC between 9.2 aquifer. Both of these processes would add dead carbon 14 and 8.5‰, as marked by letter S’’ in Fig. 7I. This range to the system and dilute CDIC. The net effect of these pro- 13 is more negative compared to the extrapolated value from cesses is the addition of DIC which has a d CDIC of 6.6‰ the regression line for all the data (i.e. 6.6‰ at point S’). (Fig. 7I). The weathering of silicate minerals is a dominant hydro- geochemical process in groundwaters with distance from 5.3. Calculation of 14C ages for groundwater Shay Gap as identified by increasing Na/Cl ratios and 87 14 13 d Sr values (Fig. 5a and b). But because we see also see When we compare CDIC and d CDIC values in a decrease in SiO2 we know that other processes such as Fig. 7III, we see two data groups emerge that also corre- clay mineral transformation reactions (Langmuir, 1997) spond with the cluster groupings derived from the hydro- must be occuring. The presence of weathering products chemistry. Groundwaters from Groups 1 and 2 have very 14 such as kaolin minerals in the core material from the Wallal low CDIC values (3.8–0.3 pmc). These waters when com- Sandstone examined at WCB04 (Drake, 1979) further sug- pared to others contain significantly higher Cl concentra- gest these processes are possible. Clay mineral processes tions. The Cl/Br ratios are higher (800–1200) than other would also explain the loss of Sr after silicate weathering groundwaters from the study site (Fig. 3a) suggesting that reactions via ion exchange reactions (Fig. 5c). If ion the primary salinity is too low to be directly related to the exchange is occurring we would see increased carbonate dis- dissolution of evaporites but that it may be contributed solution adding further dead carbon to the system (Han from a mixture of waters. The source of salinity to these et al., 2012). However, if carbonate dissolution was the only groundwaters has been hypothesised to be from mixing or process leading to the [DIC] increase in groundwater then intrusion of higher salinity water associated with basement the linear extrapolation of the regression line would extend or structural changes in the western boundary of the Basin 14 to point S in Fig. 7I. This is because the carbonate minerals (Leech, 1979). Due to low CDIC values no age calculations in the aquifer should have d13C value close to 0‰, espe- could be completed for these waters using this isotope. cially for the Wallal Sandstone which forms the basal unit With the exception of two samples (sample WCB18 and of a marine transgression (Leech, 1979). Higher carbonate PB3), groundwaters from cluster groups 4b and 4c form a 14 13 values are confirmed with the only published study to curved relationship between CDIC and d CDIC examine d13C values for carbonates of the Canning Basin, (Fig. 7III). Therefore, the statistical modelling approach which found baseline values of +1 to +2‰ (Stephens and (Gonfiantini and Zuppi, 2003; Han et al., 2014) is used to Summer, 2003). estimate 14C ages based on the following equation 494 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 k ð14C Þ¼ þ 14 ðd13C d13C Þþ ð14C Þ refer to Fig. 7III). From the regression line in Fig. 8, a slope ln DIC 1 ln E DIC ln i k k13 of 1.692 was obtained. From the slope, the values of 13 and 14 k the C ‘‘apparent decay constant” (k14 þ k13) can be þ 14 ðd13C d13C ÞðÞ 1 k ln E i 2 obtained using (Eq. (2)). 13 14 Assuming that Ci is 100 pmc (for a system evolved 14 4 1 where k14 is the C decay constant (=1.21 10 a ), k13 under open-system conditions in the unsaturated zone ie is the apparent rate constant of the reaction(s) that affect open to soil gas CO2 influx, see Appendix B), using the d13 14 d13 14 14 the CDIC value. Ci and Ci are the initial CDIC regression line, by extrapolation to CDIC = 100 pmc d13 13 and CDIC at the start of radioactive decay and any geo- (Fig. 8), we obtain d CDIC of 15‰. On the other hand, 14 d13 14 chemical processes that alter CDIC and CDIC along assuming that Ci = 75 pmc (for a system evolved under 14 13 13 with decay of C in the aquifer. d CE is the d C value closed-system conditions in the unsaturated zone, also see ?1 ‰ 14 of DIC in the aquifer at time (i.e. 6.6 ; Fig. 7I). Appendix B), by extrapolation to CDIC = 75 pmc, we 13 If the kinetics of the geochemical/physical processes that obtain d CDIC of 13.5‰. These two values are shown 14 13 affect CDIC and d CDIC can be represented by a on the regression line (Fig. 8) and are represented as open 14 pseudo-first-order process, a plot of ln( CDIC) vs. ln and closed, respectively. These points represent the range 13 13 14 (d CE d CDIC) should form a straight line, and the slope of carbon isotopic composition of DIC values before C þ k =k k 13 of the straight line (1 14 13) can be used to calculate 13, decay and increase of d CDIC. They also represent end the rate constant, of the geochemical reactions. Thus, the member A in Fig. 7. 14 k þ k 14 overall rate constant for CDIC decrease is 14 13 (e.g. Because we do not know the exact value of Ci, but we decay of 14C and dead carbon dilution of 14C) and ground- know it is between 100–75 pmc (see Appendix B), we use a water ages can be estimated by using the equation mean value of 87.5 pmc for age calculations (as represented 14 ðk þk Þ 14C ¼ 14C e 14 13 t ð Þ as partially open in Fig. 8). Assuming that Ci = 87.5 pmc, DIC i 3 14 using the regression line, by extrapolation to CDIC = 87.5 14 14 13 14 The definition of Ci is different to the initial CDIC pmc, we obtain a d CDIC of 14.1‰. Using this initial Ci 14 value ( C0) that is used in single-sample-based correction value (87.5 pmc), the calculated ages by using Eq. (3) are models (Han and Plummer, 2016). In the single-sample- given in Table 3. 14 14 4 1 based models, C0 is the initial CDIC value used in the By knowing k14 (=1.21 10 a ), k13 14 4 1 13 14 decay equation that considers only C decay rate. In the (=1.75 10 a ), d CE (=6.6‰), Ci (=87.5 pmc), 14 14 13 statistical approach, the Ci is the initial CDIC value used and d Ci (=14.1‰), the curve in Fig. 7III (i.e. depen- 14 14 13 in the equation (Eq. (3)) for both C decay and geochem- dence of CDIC on d CDIC) is calculated using Eq. (2). 14 14 ical processes that cause decrease in CDIC along with C Groundwaters that plot on or close to the calculated curve decay (e.g. dilution by dead carbon). are likely to have moved under piston-flow conditions, 14 13 13 Fig. 8 shows a plot of ln( CDIC) vs. ln(d CE-d CDIC) which relates to a traverse from Group 3 to group 4a to for the groundwater samples. All samples from Groups 1 4b to 4c groundwaters. This indicates that the geochemical and 2 plot well below the curve (also refer to Fig. 7III for environment is relatively homogenous and that along the shaded box) and are excluded from these regression calcu- groundwater flow path, the geochemical processes occurred lations. Two samples (WCB18 and PB3) from cluster group slowly in the aquifer at a reaction rate comparable to 14C 4a were also excluded from regression calculations (also decay timescales (Han et al., 2014).

14 13 13 13 Fig. 8. Plot of ln( CDIC) vs. ln(d CE-d CDIC) for the samples. The linear relationship indicates that the processes that aﬀect the C carbon isotopic composition takes place in parallel with 14C decay. K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 495

Table 3 14 Calculated ages rounded to the nearest decade using the statistical model assuming three initial CDIC values. Sample ID Age (year) 14 14 14 Ci = 87.5 pmc Ci = 100 pmc Ci = 75 pmc WCB18 50 500 0 PB3 340 790 0 PB1 540 990 20 MMCA03E 600 1050 80 WCB15Y 860 1310 340 WCB19 1280 1730 750 WCB20Y 1680 2130 1160 WCB23B 2010 2460 1490 WCB21Y 2350 2810 1830 WCB24Y 2370 2820 1850 WCB25X 3460 3920 2940 WCB25Y 3730 4180 3210 Cooragoora PB1 4020 4480 3500 WCB17 4710 5160 4190 WCB22 5700 6150 5180 Pardo station 5850 6300 5330

13 5.4. Discussion on the statistical approach and sensitivity would result in a higher d C(CO2) value (Appendix B: testing 18.4‰). Therefore, it seems that (1) the estimated 14C 14 ages of the groundwaters using Ci = 75 are more reliable The degree of openness with respect to soil CO2 influx (i.e. the oldest age of the water collected at Pardo Station is can be assessed using the statistical approach. Our calcula- ca. 5.3 ka), and (2) the recharge of groundwater was a rel- tions show (Appendix B) that for the studied system, the atively fast process, a situation when water moves rapidly 14 Ci value would be 100 pmc assuming an open system, without carbon exchange between soil CO2 and DIC in and 75 pmc assuming a closed system. If the system was the unsaturated zone (Han and Plummer, 2016). 14 partially open to soil CO2, the value of Ci would be Based on the radiocarbon dating results, groundwaters between 75 and 100 pmc. Therefore, the uncertainty of that are located inland where the confining layer of the Jar- age estimation using this method depends on the assumed lemai Siltstone is absent (Fig. 1b), are modern near PB1 and 14 Ci. PB3, they increase to 1.5 ka (WCB23) and increase along To test the sensitivity of the statistical approach to the the groundwater flow path to 5.3 ka (Pardoo Station) 14 selection of Ci, the ages are calculated by using 100 and near the coast (Table 3). It was calculated that it would take 75 pmc values, respectively, in addition to 87.5 pmc. An between 0.2 and 15 ka to travel across the site when calcu- 14 uncertainty of ±12.5 pmc in estimated and Ci causes an lating an average linear velocity based on a hydraulic gradi- age uncertainty of 15.4% for the oldest water (of which ent of 0.7 and the large variation in hydraulic conductivity the upper age limit obtained from the statistical approach measured in the Wallal Sandstone ranging from 1 to is 6.3 ka; Table 3). 100 m day1 (Leech, 1979). Particle tracking travel times The initial isotopic composition of DIC (end member A) of approximately 5 ka to travel across the model domain 13 may have values between 15 and 13.5‰ for d CDIC, in the west-south-west groundwater flow direction were cal- 14 between 75 and 100 pmc for CDIC, depending on how culated (Aquaterra, 2010) and agree with the radiocarbon open it is to soil CO2 influx (Appendix B). The calculations ages. Please note this excludes samples from zones 1 and 2. 13 show that the soil CO2 should have d C(CO2) between There are several advantages of using the statistical 18.4 and 16.9‰. These values suggest soil CO2 has orig- model. One advantage is that an estimate of radiocarbon inated from a mixture of C3 and C4 vegetation types which age can be made without knowing the exact soil gas 13 seems reasonable considering the surface vegetation of this d CCO2(g) value. Estimating this value in a mixed C3 and arid region is made up of a mixture of Mulga, woodlands C4 vegetation environment is challenging if the actual soil and shrublands with hummock grasslands (McFarlane, gas was not measured. Even if the soil gas is measured, it 2015). The only study to measure the soil CO2 from the may be difficult to estimate the relative contribution of C3 unsaturated zone in an arid region in Central Australia, and C4 vegetation under varied climates throughout time. 13 found a narrow range of d CCO2 values between 13.5 The other advantage is that in systems where the DIC has and 16.8‰, which they suggested resulted from spinifex undergone complex geochemical reactions, the age can be grass respiration (Wood et al., 2014). If we consider these estimated by using the statistical approach, provided that values, we can infer that the system studied has evolved the processes can be lumped into a pseudo-first-order pro- under closed-system conditions with respect to soil CO2, cess (Han et al., 2014). For such complex geochemical sys- 13 14 (ie Appendix B: d C(CO2) = 16.9‰ and Ci = 75 pmc). tems the single-sample-based models may result in In contrast, a system evolved under open-system conditions significantly biased or completely wrong results. 496 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498

6. CONCLUSION the coast. The groundwater age calculations compared well with particle tracking modelling except for groundwaters The study was undertaken to assess the groundwater located in the western section (Groups 1 and 2) which are recharge sources and calculate groundwater age for the of a different origin. The baseline data collected and anal- Wallal Sandstone. To the authors knowledge this study ysed from this study can be used alongside time-series data provides the first groundwater isotope database and age to help assess the sustainability of large groundwater basins calculations for Australia’s second largest groundwater into the future. basin, the Canning Basin. The negative d18O and d2H values of groundwaters within the study site suggest that they ACKNOWLEDGEMENTS have been recharged by precipitation originating from tropical cyclones. Water recharges the soil zone rapidly after The authors would like to thank the Government of Western these events and the groundwater undergoes silicate mineral Australia (WA), Department of Water (DoW) for providing the funding for the field work and the cost of analysis. A special thank weathering and clay mineral transformation processes. Par- you to Hazli Koomberi for his enthusiasm and support throughout tial carbonate dissolution processes occur within the satu- the field work. The writing and data analysis was funded by rated zone, but under closed system conditions. As the ANSTO. We would also like to thank ANSTO analytical and tech- 14 groundwater moves, C-free DIC is added continuously nical staff in providing the hydrochemical and isotope results: in to the system caused by complex processes, and on a similar particular Alan Williams, Simon Varley, Vladimir Levchencko, time scale to radiocarbon decay. Andrew Jenkinson, Barbara Neklapilova, Henri Wong, Robert The groundwaters that displayed a curved relationship Chisari and Kelly-Anne Farrawell. 14 13 between CDIC and d CDIC, groundwater ages were calculated using the statistical approach. A test of the sensitivity 14 of this method to the selection of Ci was undertaken and APPENDIX A. METHODS OF GROUNDWATER it was found that 15% error could be found for the oldest ANALYSIS USED IN THIS STUDY water. Groundwater age was found to increase from where the Wallal Sandstone was unconfined in the south east to around 5.3 ka along the groundwater flow path towards

Analyte Bottle/preservation Analysis Units/reference Error Reference material Cations 60 ml acid washed HDPE/ Inductively coupled plasma – atomic mg/L Meredith HNO3 emission spectroscopy et al. (2012) Anions 60 ml HDPE Ion chromatography mg/L Meredith et al. (2012) d18O, 30 ml HDPE Isotope mass spectrometry (IRMS) Per mil (‰)/V-SMOW ±0.1, ±1‰ Meredith d2H (Vienna Standard Mean et al. Ocean Water) (2009) 13 13 d CDIC 12 ml exetainer combusted IRMS ‰/NBS19 (d C = +1.95 ±0.1‰ Meredith @500 °C for 4 h VPDB) et al. (2016) 87Sr/86Sr 60 ml HDPE acid washed Finnigan MAT-261 thermal ionisation NBS-981 (of 0.71020) ±0.00002 X3 with double sub-boiled mass spectrometer. Mass fractionation pure HNO3/preserved with during the analysis was corrected by this acid normalising the isotopic compositions to 86Sr/88Sr = 0.1194 14 CDIC 1000 ml HDPE Accelerator mass spectrometry Percent Modern Carbon average 1r error Meredith (pMC), converted to of ±0.20 et al. pmc according to (2012) Meredith et al. (2016) 3H 1000 ml HDPE Liquid scintillation counter Tritium units (TU) Uncertainty = Meredith ±0.1 TU, et al. quantiﬁcation (2012) limit = 0.4 TU K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498 497

14 APPENDIX B. CALCULATIONS OF INITIAL Calculation results: Ci = 0.5 100 + 0.5 50 = 75 13 CARBON ISOTOPIC CONTENTS pmc; d Cg = 16.9‰.

Assuming that the pH value of the water is 6.4, REFERENCES and the temperature is 25 °C. In diluted aqueous solutions, at a temperature of 25 °C and pH 6.4, the mole Aquaterra, 2009. West Canning Basin Model Design. May 2009. No 1033/B/021c. fraction of CO2(aq) and HCO3(aq) are equal, i.e. [CO2(aq)]: Aquaterra, 2010. West Canning Basin Modelling Project. May [HCO3(aq)] = 1:1. n n 14 2010. No. 1033B Report 049a. In the following calculations, the Ci values are calcu- 14 BOM, 2013. Climate statistics for Australian sites. Commonwealth lated first. The calculated results are Ci = 100 pmc (for of Australia. Bureau of Meteorology (ABN 92 637 533 532). 14 a completely open system), and Ci = 75 pmc (for a com- BOM, 2016. http://www.bom.gov.au/cyclone/history/wa/roe- 14 pletely closed system). Assuming that Ci = 100 pmc, using bourne.shtml. 14 the regression line, by extrapolation to CDIC = 100 pmc Dogramaci S. and Skrzypek G. (2015) Unravelling sources of 13 (Fig. 8), we obtain d CDIC of 15‰. On the other solutes in groundwater of an ancient landscape in NW 14 Australia using stable Sr, H and O isotopes. Chem. Geol. hand, assuming that Ci = 75 pmc, by extrapolation to 14 13 393–394, 67–78. CDIC = 75 pmc, we obtain d CDIC of 13.5‰. These two value pairs are shown in Fig. 8 on the regression line. Dogramaci S., Skrzypek G., Dodson W. and Grierson P. F. (2012) Stable isotope and hydrochemical evolution of groundwater in These points represent the range of the carbon isotopic 14 the semi-arid Hamersley Basin of subtropical northwest Aus- composition of DIC before C decay, representing tralia. J. Hydrol. 475, 281–293. d13 end member A in Fig. 7I. These two CDIC values Drake, J.R., 1979. Identification of 4 rocks from West Canning (i.e. 15‰ and 13.5‰) are then used in the following Basin Bores. Petrology report No 647. Western Australia, equations, on the right hand side. Geological Survey, Perth, March 1979. 13 13 13 Definition: d Cg is soil CO2 d C value. da/g is the C Elderfield H. (1986) Strontium isotope stratigraphy. Palaeogeogr. fractionation factor of CO2(aq) with respect to gaseous soil Palaeoclimatol. Palaeoecol. 57, 71–90. 13 Faure G. and Mensing T. M. (2005) Isotopes: Principles and CO2, and db/g is the C fractionation factor of HCO3(aq) applications (3rd edition). John Wiley & sons Inc, Hoboken, with respect to gaseous soil CO2. New Jersey, USA, p. 897. Fontes J. C. F. and Ch J. (1980) Environmental Isotopes in (1) Evolution of DIC (i.e. end member A) under Groundwater Hydrology. Elsevier, Netherlands. open-system conditions. Fudeyasu H., Ichiyanagi K., Sugimoto A., Yoshimura K., Ueta A., Yamanaka M. D. and Ozawa K. (2008) Isotope ratios of This is the situation when water moves slowly in the precipitation and water vapor observed in Typhoon Shanshan. unsaturated zone to allow carbon exchange between DIC J. Geophys. Res.: Atmos. 113, n/a-n/a. Gedzelman S., Lawrence J., Gamache J., Black M., Hindman E., and soil CO2 until equilibration. Black R., Dunion J., Willoughby H. and Zhang X. (2003) þ ¼ CO2ðaqÞ HCO3ðaqÞ DIC Probing hurricanes with stable isotopes of rain and water mole fraction 0:50:51 vapor. Mon. Weather Rev. 131, 1112–1127. 13 13 13 Gleeson T., Befus K. M., Jasechko S., Luijendijk E. and Cardenas d CDIC d Cg þ ea=g d Cg þ eb=g 15 M. B. (2016) The global volume and distribution of modern 14C 100 100 14C DIC i groundwater. Nat. Geosci. 9, 161–167. 14 where, da/g = 373/T + 0.19 = 1.06‰, and db/g = Geyh M. A. (2000) An overview of C analysis in the study of groundwater. Radiocarbon 42, 99–114. dg/b = [9483/T + 23.89] = 7.92‰ 14 Gleeson T., Wada Y., Bierkens M. F. P. and van Beek L. P. H. Calculation results: Ci = 0.5 100 + 0.5 100 = 100 pmc; d13C = 18.4‰ (2012) Water balance of global aquifers revealed by ground- g water footprint. Nature 488, 197–200. Gonfiantini R. and Zuppi G. M. (2003) Carbon isotope exchange (2) Evolution of DIC (i.e. end member A) under closed- rate of DIC in karst groundwater. Chem. Geol. 197, 319–336. system conditions. Gonfiantini, R., 1972. Notes on isotope hydrology. Internal report. Vienna, IAEA. This is the situation when water moves rapidly without Haig, T., 2009. The Pilbara Coast Water Study, Department of Water, Hydrogeological record series, Report No. HG34, 173 carbon exchange between DIC and soil CO2 in the unsaturated zone. Assuming that carbonate minerals contains no pp. 14C, and the d13C value of the HCO produced from dis- Han L.-F., Niel Plummer L. and Aggarwal P. (2014) The curved 3(aq) 14C vs. d13C relationship in dissolved inorganic carbon: A solution of carbonate is 0‰. Thus, half of HCO 3(aq) useful tool for groundwater age- and geochemical interpreta- (resulted from carbonate minerals) contains only dead car- tions. Chem. Geol. 387, 111–125. 14 bon, and has carbon isotopic composition of CDIC =0 Han L.-F., Plummer L. N. and Aggarwal P. (2012) A graphical 13 and d CDIC =0. method to evaluate predominant geochemical processes occurring in groundwater systems for radiocarbon dating. Chem. CO ð Þ þ HCO ¼ DIC 2 aq 3ðaqÞ Geol. 318–319, 88–112. mole fraction 0:50:51Han L.-F. and Plummer L. N. (2013) Revision of Fontes & 13 13 1 13 14 d d þ e = ðd þ e = Þ: Garnier’s model for the initial C content of dissolved CDIC Cg a g 2 Cg a g 13 5 Chem. Geol. 14C 100 50 14C inorganic carbon used in groundwater dating. DIC i 351, 105–114. 498 K.T. Meredith et al. / Geochimica et Cosmochimica Acta 220 (2018) 483–498

Han L. F. and Plummer L. N. (2016) A review of single-sample- Puls, R., Barcelona, M., 1996. EPA ground water issue: low-flow based models and other approaches for radiocarbon dating of (minimal drawdown) ground-water sampling procedures; EPA/ dissolved inorganic carbon in groundwater. Earth Sci. Rev. 152, 540/S-95/504. 119–142. Plummer, L.N., Glynn, P.D., 2013. Radiocarbon dating in Hartigan J. A. and Wong M. A. (1979) A K-means clustering groundwater systems, in Isotope Methods for Dating Old algorithm. J. Royal Stat. Soc. Ser. C (Appl. Stat.) 28, 100–108. Groundwater. International Atomic Energy Agency, Vienna, Ingerson E. and Pearson F. J. (1964) Estimation of Age and Rate of April, 2013, Chap. 4, pp. 3–89. STI/PUB/1587, ISBN 978–92– Motion of Groundwater by the 14C-method. Recent Researches 0–137210–9, 357 p. http://www-pub.iaea.org/books/IAEA- in the Fields of Atmosphere, Hydrosphere, and Nuclear Geo- Books/8880/Isotope-Methods-for-Dating-Old-Groundwater. chemistry, Sugawara Festival Volume. Maruzen Co, Tokyo. Richey A. S., Thomas B. F., Lo M.-H., Reager J. T., Famiglietti J. Jollife I. (1986) Principal Component Analysis. Springer, New York. S., Voss K., Swenson S. and Rodell M. (2015) Quantifying Langmuir D. (1997) Aqueous Environmental Geochemistry. Prentice renewable groundwater stress with GRACE. Water Resour. Hall, Upper Saddle River, New Jersey. Res. 51, 5217–5238. Lawrence J. R., Gedzelman S. D., Gamache J. and Black M. (2002) Scanlon B. R. (2000) Uncertainties in estimating water fluxes and Stable isotope ratios: Hurricane Olivia. J. Atmos. Chem. 41, 67– residence times using environmental tracers in an arid unsat- 82. urated zone. Water Resour. Res. 36(2), 395–409. Lawrence R. J. and Gedzelman D. S. (1996) Low stable isotope Semeniuk V. (1996) Coastal forms and quaternary processes along ratios of tropical cyclone rains. Geophys. Res. Lett. 23, 527–530. the arid Pilbara coast of northwestern Australia. Palaeogeogr. Leech, R.E.J., 1979. Geology and groundwater resources of the Palaeoclimatol. Palaeoecol. 123, 49–84. Southwestern Canning Basin, Western Australia. Geological Stephens N. P. and Summer D. Y. (2003) Late Devonian carbon Survey of Western Australia. Record 1979/9. Perth 1979. ISBN isotope stratigraphy and sea level fluctuations, canning Basin, No. 072448034X. Western Australia. Palaeogeogr. Palaeoclimatol. Palaeoecol. McFarlane, D., 2015. Pilbara Water Resource Assessment. A 191, 203–219. report to the Government of Western Australia and industry Sturman A. and Tapper N. (1999) The Weather and Climate of partners from the CSIRO Pilbara Water Resource Assessment. Australian and New Zealand. Oxford University Press, Oxford. CSIRO Land and Water, Australia. McFarlane DJ (ed.). Taylor R. (2014) Hydrology: When wells run dry. Nature 516, 179– Meredith K., Cendoń D. I., Pigois J.-P., Hollins S. and Jacobsen G. 180. (2012) Using 14C and 3H to delineate a recharge ‘window’ into Taylor R. G., Scanlon B., Doll P., Rodell M., van Beek R., Wada the Perth Basin aquifers, North Gnangara groundwater system, Y., Longuevergne L., Leblanc M., Famiglietti J. S., Edmunds Western Australia. Sci. Total Environ. 414, 456–469. M., Konikow L., Green T. R., Chen J., Taniguchi M., Bierkens Meredith K. T., Han L. F., Hollins S. E., Cendoń D. I., Jacobsen M. F. P., MacDonald A., Fan Y., Maxwell R. M., Yechieli Y., G. E. and Baker A. (2016) Evolution of chemical and isotopic Gurdak J. J., Allen D. M., Shamsudduha M., Hiscock K., Yeh composition of inorganic carbon in a complex semi-arid zone P. J. F., Holman I. and Treidel H. (2013) Ground water and environment: Consequences for groundwater dating using climate change. Nat. Clim. Change 3, 322–329. radiocarbon. Geochim. Cosmochim. Acta 188, 352–367. Venables, W., Smith, D., R.C., 2016. An introduction to R. https:// Meredith K. T., Hollins S. E., Hughes C. E., Cendoń D. I., Hankin cran.r-project.org/doc/manuals/r-release/R-intro.pdf. S. and Stone D. J. M. (2009) Temporal variation in stable Wood C., Cook P. G., Harrington G. A., Meredith K. and Kipfer isotopes (18O and 2H) and major ion concentrations within the R. (2014) Factors affecting carbon-14 activity of unsaturated

Darling River between Bourke and Wilcannia due to variable zone CO2 and implications for groundwater dating. J. Hydrol. ﬂows, saline groundwater inﬂux and evaporation. J. Hydrol. 519(Part A), 465–475. 378, 313–324. Wright W. J. (1997) Tropical–extratropical cloudbands and Aus- Munksgaard N. C., Zwart C., Kurita N., Bass A., Nott J. and Bird tralian rainfall: I. climatology. Int. J. Climatol. 17, 807–829. M. I. (2015) Stable isotope anatomy of tropical cyclone Ita, https://doi.org/10.1002/(SICI)1097-0088(19970630)17:8<807:: North-Eastern Australia, April 2014. PLoS One 10, e0119728. AID-JOC162>3.0.CO;2-J. Parkhurst D. and Appelo C. (1999) User’s guide to PHREEQC Zwart C., Munksgaard N. C., Kurita N. and Bird M. I. (2016) (version 2) – a computer program for speciation, batch- Stable isotopic signature of Australian monsoon controlled by reaction, one dimensional transport, and inverse geochemical regional convection. Quatern. Sci. Rev. 151, 228–235. calculations. USGS Water Resour. Investigations Rep. 99, 4259. Associate editor: Orit Sivan