Groundwater residence time in a dissected and weathered sandstone plateau: Kulnura–Mangrove Mountain aquifer, NSW, Australia

Groundwater residence time in the Kulnura–Mangrove Mountain aquifer was assessed during a multi-year sampling programme using general hydrogeochemistry and isotopic tracers (H2O stable isotopes, δ13CDIC, 3H, 14C and 87Sr/86Sr). The study included whole-rock analysis from samples recovered during well construction at four sites to better characterise water–rock interactions. Based on hydrogeochemistry, isotopic tracers and mineral phase distribution from whole-rock XRD analysis, two main groundwater zones were differentiated (shallow and deep). The shallow zone contains oxidising Na–Cl-type waters, low pH, low SC and containing 3H and 14C activities consistent with modern groundwater and bomb pulse signatures (up to 116.9 pMC). In this shallow zone, the original Hawkesbury Sandstone has been deeply weathered, enhancing its storage capacity down to ∼50 m below ground surface in most areas and ∼90 m in the Peats Ridge area. The deeper groundwater zone was also relatively oxidised with a tendency towards Ca–HCO3-type waters, although with higher pH and SC, and no 3H and low 14C activities consistent with corrected residence times ranging from 11.8 to 0.9 ka BP. The original sandstone was found to be less weathered with depth, favouring the dissolution of dispersed carbonates and the transition from a semi-porous groundwater media flow in the shallow zone to fracture flow at depth, with both chemical and physical processes impacting on groundwater mean residence times. Detailed temporal and spatial sampling of groundwater revealed important inter-annual variations driven by groundwater extraction showing a progressive influx of modern groundwater found at >100 m in the Peats Ridge area. The progressive modernisation has exposed deeper parts of the aquifer to increased NO3− concentrations and evaporated irrigation waters. The change in chemistry of the groundwater, particularly the lowering of groundwater pH, has accelerated the dissolution of mineral phases that would generally be inactive within this sandstone aquifer triggering the mobilisation of elements such as aluminium in the aqueous phase.


INTRODUCTION
Chemical weathering by infiltrating freshwater in sandstone can result in drastic changes to the porosity and permeability properties of the original material (Emery et al. 1990;Liu et al. 1996;França et al. 2003). With sufficient rainfall, the extent of weathering will depend on the original mineralogy, sedimentary facies, diagenetic history and geological setting of the aquifer materials.
The relevance of weathering changes is well understood by petroleum geologists as they characterise the reservoir capacity of formations (e.g. Ali et al. 2010;Morad et al. 2010;Taylor et al. 2010, among many others). However, its impact on groundwater recharge, flow and groundwater age has attracted little attention in hydrological studies.
The concept of groundwater age has evolved since the first applications to groundwater research (Muennich 1957), with the influence of hydrodynamic dispersion (heterogeneity and groundwater mixing) being recognised as causing important changes in tracer concentrations (Varni & Carrera 1998;Weissmann et al. 2002, among others). Groundwater age can be better expressed as a mean of the age distribution within a sample (Bethke & Johnson 2008), and only when approaching a piston flow model do tracer groundwater residence times closely represent the distribution of ages present in the sample.
The dating range of radiocarbon ( 14 C) fits well within the recharge time frame of many aquifers worldwide and covers a time gap spaning important climatic variations during the Late Pleistocene and Holocene, capturing modified rainfall regimes and thus groundwater recharge processes. While the upper dating limit of modern 14 C techniques (with appropriate sample treatment) can be close to ca 60 ka in well-constrained systems (without C exchange) (Turney et al. 2001), limitations imposed by potential interactions of groundwater and its dissolved carbon in the aquifer matrix curtail its realistic potential dating range to ca 30 ka. Apart from these chemical processes, physical factors such as diffusion can further impact 14 C concentrations particularly in fractured systems (Sanford 1997;Cook et al. 2005). In order to account for these chemical and physical dilution effects, a number of geochemical and hydrogeological models can be applied producing sometimes different estimations of groundwater residence times (Clark & Fritz 1997). The discussion of 14 C geochemical correction models is beyond the scope of this paper, and the reader is referred to the many detailed papers available (Gallagher et al. 2000;Coetsiers & Walraevens 2009;Cartwright 2010;Han et al. 2012, among many others).
The Hawkesbury Sandstone and underlying Narrabeen Group lithologies formed in varied depositional environments with changes in sediment sources during deposition and precipitation of dispersed cementing carbonates resulting in relatively heterogenous units. Shale lenses observed throughout the Hawkesbury Sandstone can lead to the development of perched groundwater systems that in turn enhance weathering of overlying materials (Loughnan et al. 1962). Intense local weathering aided by fracture networks and shale lenses has facilitated porosity and permeability changes, with groundwater flow regimes changing from initially fracture controlled to semi-porous particularly in shallow areas. The result of these various depositional and weathering processes is the formation of a complex groundwater system where the areal extent and depth of deeply weathered materials are variable, reaching up to $90 m, as observed in areas of the Sydney Basin (Pecover 1986;Freed 2005). The extensive weathering favours the development of shallow groundwater systems, which in turn support productive agricultural areas in the Sydney Basin.
In the Kulnura-Mangrove region, groundwater extraction for potable water supply and for industrial activities can coexist, provided the main recharge areas are protected, pumping does not exceed recharge, and the basic parameters within the aquifer are known through appropriate studies. Effective use of environmental isotopes has been demonstrated for identifying recharge areas (Meredith et al. 2012), groundwater mixing processes (Cend on et al. 2010;Cartwright et al. 2012;Currell et al. 2013) and tracing groundwater responses to potential anthropogenic changes such as pumping, pollution, subsidence effects after coal mining or coal-seam gas drilling (Frost et al. 2002;Osborn et al. 2011;Payne et al. 2013).
In this paper, we examine the links between deep weathering, groundwater residence time and anthropogenic changes (e.g. pumping, agricultural inputs); comparing groundwater from three sandstone plateaus with similar geological characteristics but variable weathering depths and levels of groundwater extraction. We apply several hydrogeochemical and isotopic techniques to groundwater ages, constraining the potential factors affecting 14 C in groundwater, particularly in shallow groundwater, with the use of other isotopes such as 3 H, H 2 O-d 2 H and d 18 O, and 87 Sr/ 86 Sr. The general aims are: (i) to determine the main sources of solutes including those derived from water-rock reactions that can affect 14 C activity (a 14 C), (ii) to characterise groundwater residence times in the Kulnura-Mangrove Mountain aquifer (KMMA) comparing groundwater ages in the three main sandstone plateaus, (iii) to characterise the effects of deep weathering on groundwater residence times, and (iv) to assess potential effects on groundwater from anthropogenic activities such as pumping and agriculture.

Physiography and water usage
The Kulnura-Mangrove Mountain (KMM) plateau is located approximately 50 km north of Sydney (NSW, Australia) and 20 km from the coast, inland from the town of Gosford (Figure 1). The area forms an elevated plateau or tableland with a maximum elevation at the northern end of 386 m Australian Height Datum (AHD). The plateau consists of gently undulating land bounded by rugged cliffs and incised valleys that define subparallel drainage along the Mooney Mooney, Popran and Mangrove creeks, incising to near sea-level and flowing into the Hawkesbury River.
Drainage from this region is partly captured by water storage (i.e. Mangrove Mountain Dam) that contributes to municipal water for the Gosford-Wyong Area. Approximately 10% of land use is allocated as a drinking-water reserve with $40% of the land area covered by National Parks and State Forests, with a large portion of the flat to gently sloping land areas cleared for agricultural and industrial pursuits. The plateau comprises predominantly deeply weathered and fractured sandstone of the Hawkesbury Sandstone that hosts the KMMA covering $506 km 2 (Bish et al. 1994). Groundwater use is managed by the NSW Office of Water according to the Water Sharing Plan for the KMM Groundwater Sources 2003(NSW DPNIR 2004. Approximately 6 GL of groundwater have been authorised for extraction with around 70% taken for commercial activities including poultry farming, mineral water bottling, extractive industries (quarries), nurseries and citrus irrigation. Without a reticulated water supply available throughout much of the area, groundwater is also an important water resource to the local communities for domestic and stock purposes accounting for approximately 30% of the total licensed extraction.

Geology
The outcropping rocks in the KMMA study area are predominantly middle Triassic in age and comprise part of the uppermost units of the Sydney Basin sequence, the Hawkesbury Sandstone and underlying Terrigal Formation. These sequences are intruded in places by Jurassic diatremes and several episodes of Paleogene volcanic flows (Crawford et al. 1980;Embleton et al. 1985) ( Figure 2). The Permian-Triassic strata dip gently to the south-southwest, towards the centre of the Sydney Basin with a series of north-south-trending broad fold-like structures imposed on this regional dip. The axes of these folds plunge very gently to the south-southwest. In the KMMA, the folds are the Yarramalong Syncline, along the eastern boundary of the plateau, and the Kulnura Anticline, along the western boundary. Many largescale fracture traces and lineaments in four key orientations have been mapped as intersecting the study area (Mauger et al. 1984). At least one of these features has been related to a series of 10 m offset normal faults and an associated graben structure encountered during the construction of the Boomerang Creek tunnel (NSW Department of Public Works 1991), which transects the north of the study area.
On a smaller scale, widely spaced, mostly planar, vertical joint sets are common in all the massive and bedded sandstone units. In the shale lenses, jointing is usually more closely spaced and variable than in the adjacent sandstones. Generally, joints do not extend across the bedding from sandstone, through shale, to adjacent sandstone beds. Individual joints are not continuous for more than 15-30 m vertically, and 100 m horizontally. Spacing between major parallel joints averages 10 m in the Hawkesbury Sandstone, with inclinations ranging 20 either side of vertical (Chesnut 1988). Near the study site, the orientation of the major joint direction is to the northeast, with a complementary set to the northwest (Longworth & MacKenzie 1985).

TERRIGAL FORMATION
The Terrigal Formation is the uppermost unit of the Narrabeen Group and underlies the Hawkesbury Sandstone. The upper Terrigal Formation outcrops in drainage valley walls and on all sides of the study area plateau in Mangrove Creek and Wyong Creek. The Terrigal Formation consists of an interbedded sequence of sandstone, siltstone and claystone some 200 m thick. Bedding in the upper Terrigal Formation is typically 5-10 m thick and is separated by thin shale layers a few metres apart and laterally persistent for over 300 m. In general, the sandstone units of the Terrigal Formation have a smaller average grainsize, more clay and silt matrixand hence lower primary porositycompared with the Hawkesbury Sandstone. They also tend to have   (Uren 1974;Nicholas Papua & Associates 1976;Longworth & McKenzie 1985;NSW DPI 2009). The three cross-sections show the screened intervals (to scale), and numbers represent the different piezometers. The inferred gradual transition between the Hawkesbury Sandstone (Rh) and the Terrigal Formation is represented by a shaded area, with the band corresponding to the range of available regional information. All cross-sections contain a graphic representation of proportions of siderite and clay minerals found by XRD analysis of whole rock, and the sharper peaks correspond to shale lenses (Table S2). The cross-section also shows the average groundwater residence times (year BP). carbonate cement, rather than silica or kaolinite, and lack overgrowths on quartz grains (McNally 1995).

HAWKESBURY SANDSTONE
The gently dipping (1-2 S) Hawkesbury Sandstone is around 100 m thick with 94 m intersected at Kulnura in the north (Bradley & Stuntz 1964) and 96 m at Somersby in the southern edge of the study area (Nicholas Papua & Associates 1976). Further south, 290 m is observed at the Hawkesbury River; the maximum thickness recorded anywhere in the Sydney Basin (Herbert 1983).
The Hawkesbury Sandstone is composed mainly of medium-coarse sized quartz grains within a silica-siderite cement and a variable proportion of clay and rock fragment matrix. Average estimates of the clay matrix content ranges from 20 to 40% (Standard 1964;Bowman 1974). Individual sandstone beds usually occur as elongated lenses typically up to 2 m thick and 300 m long, but in places up to 10 m thick and 1 km long (Standard 1964). Distributed within the sandstone, shale lenses have been mapped, ranging from a few centimetres to $7 m in thickness, but have a limited areal extent (<0.08 km 2 ), resulting in localised aquitards. The shale lenses are terminated laterally by becoming gradually more interbedded with sandstone, or by being truncated by a channel sandstone or a scour and fill zone (Uren 1974). Generally, beds in the Hawkesbury Sandstone have been classified into three facies related to depositional conditions with details summarised in Table 1.

DEEP WEATHERING
Parts of the KMM plateau show evidence of deep weathering: deposits of soft and friable sandstone have developed on the top of the Hawkesbury Sandstone, some of which are quarried for construction sand. These friable sandstones generally exhibit higher porosities than harder sandstones and are proposed to behave as zones of rapid recharge into the KMMA.
The formation of friable sandstone is thought to be the result of deep weathering along extensive fractures and numerous perched water-tables that concentrate the downward percolation of large volumes of water; dissolving, leaching and redepositing the cementing material (Pecover 1986). In between zones of friable sands, harder unaltered rock persists forming rocky knolls, rising above surrounding masses of soft deeply weathered friable sandstone. Outcropping sandstone tends to become case hardened owing to surface effects of dissolution and re-cementing of iron and silica, which further aids resistance to erosion and can lead to the formation of silica karst features (Pecover 1986;Young & Wray 2000). The depth and extent of friable sandstone have been accurately defined only where quarried or under areas proposed for quarrying. The depth of friable sandstone, in the proposed southern extension of Calga Sand Quarry, was reported to vary between 20 and 30 m below ground level (R. W. Corkery 2009). The maximum depth of friable sandstone at the proposed Somersby Fields project was reported at 23 m ( RCA Australia 2006). At both locations, underlying the friable sandstone is an indurated zone that is non-rippable for the purpose of sand quarrying. The extent and distribution of friable sandstone over the remainder of the study area are not well known, although its widespread presence has been inferred from the successful development of agriculture over the plateau and the dominantly high yield of groundwater wells above 30 m depth. Deep weathering reaching $90 m is deduced from mineralogical studies within the Mangrove central plateau. Weathered sandstone reaching depths of up to 100 m have been observed by drilling in the Newnes Plateau within Narrabeen Group sandstones on the western side of the Sydney Basin (Pecover 1986). In both cases, the area of deeper weathering is bounded by major lineaments. Similar Table 1 Summary of Hawkesbury Sandstone facies characteristics with hydrological characteristics (Conaghan 1980;Rust & Jones 1987).

Facies
Sedimentary Structure Lithology Hydro characteristics 'Massive' Little or no structure; typically have an erosional base.
Very poorly sorted, fine-medium size sands, dispersed gravels and claystone fragments.
Reduced permeability owing to poor sorting, lack of structure and higher amounts of distributed clay and rock fragments. Lower amounts of cement and lower primary macro-porosity compared with Sheet Facies. 'Sheet' or 'Stratified' Large-scale foreset cross-beds occurring in sets and inclined at $20 to the NE. Typically with a conforming base.
Medium to very coarse sand grains bound by silica cement with minor siderite.

Climate
The KMM area has a warm temperate climate with maritime influences. There are prolonged periods of high temperature, particularly during summer, and also periods of high rainfall. The mean annual rainfall at Peats Ridge (BOM 2013) was 1257 mm for the period 1980-2012 ( Figure 3). The monthly average rainfall varies between 66.7 mm in July and 159.3 mm in February. The annual average pan evaporation is 1544 mm. Most significant rainfall events in winter in this region involve air masses that have travelled from the Tasman Sea pushed by low-pressure systems. Inland troughs bringing moisture from the northeast also result in periodic large rainfall events. Alternating dry and wet periods are common and can be prolonged as they are linked to wider climatic events such as El Niño and La Niña, and modulated by the Indian Ocean Dipole (Ummenhofer et al. 2009). The region has experienced one wet period from 1985 to 1990 between the dry periods before 1985 and 1991-2007 as indicated by cumulative rainfall residual (Figure 4).

Hydrogeology
The Hawkesbury Sandstone and units of the underlying Narrabeen Group, are generally considered to behave as a fractured rock aquifer system with dual porosity: where the primary porosity of the rock pore space is dominated by a secondary porosity created by weathering, jointing, fractures, cross-bedding and separations between beds (McKibbin & Smith 2000). Joints are the most common rock defects and usually 'opened' by stress relief (unloading) owing to erosion overhead or in adjacent valleys (McNally 1981). Conversely, joints are generally considered to become increasingly watertight with depth with a corresponding reduction in aquifer permeability. Packer test studies conducted basin-wide indicate an order of magnitude reduction in hydraulic conductivity with depth for the Hawkesbury Sandstone over the top 100 m (Tammetta & Hawkes 2009).
Several conceptual models have been proposed for groundwater distribution in the KMMA (Williams 1984;Bish et al. 1994;Alkhatib & Merrick 2006;Cook 2010). The central premise of all models is that the accessible groundwater is contained in a series of stacked subaquifers distributed throughout the Hawkesbury Sandstone. The subaquifers are proposed to be perched above shale layers or above permeability contrasts between porous, coarse-grained cross-stratified sandstone beds (sheet facies) and tightly cemented or clay-rich beds (massive facies). Both styles of aquitards, shales and massive facies, are known to be laterally discontinuous, resulting in local-scale variability in aquifer distribution both horizontally and vertically. The aquifers are also connected to varying degrees by complementary sets of subvertical joints and fractures.
The ridge topography has the effect of dewatering aquifers towards the margin of the valleys, and in general, shallow aquifers are not found at these locations. Hydraulic gradients must be considered as a two-layered system. The upper layer is strongly affected by topography with the hydraulic gradient toward the valleys. The lower layer, near or below the base of valleys, is controlled by the regional dip of the Hawkesbury Sandstone and plateau with the hydraulic gradient towards the south.
There are numerous production wells in the area ranging in depth from 16 to 130 m with most concentrated in the upper 60 m. Williams (1984) recognised a correlation between height of topographic surface and depth of the wells indicating a possible major single aquifer. Generally, groundwater yields range between 0.1 and 4.5 LÁs À1 with higher yields obtained at depths <30 m. Transmissivity from open production wells ranges from $5 m 2 Ád À1 to 50 m 2 Ád À1 (Williams 1984).   Table S1 for well characteristics and Figure 1 for location. Missing data correspond to periods of instrument failure. All water-level data are from NOW. drought in 2007 ( Figure 4). The exception to natural water-level variations can be observed in well 271008-1/2 ( Figure 4b, c) where natural trends are overprinted by sharp and continuous variations in groundwater levels caused by pumping. Despite pumping, groundwater levels tend to recover following rainfall.

Rock fragments and soil sampling
Whole-rock powder samples were collected from wells 271007-3, 271008-3, 271009-3 and 271012-3 during construction, with detailed isotopic analysis carried out for 271007-3. All wells reached variable depths from 50 m to a maximum of 150 m (Supplementary Information  Table S1). Powdered samples from 271007-3 were first analysed with a LECO (CNS2000) elemental analyser to determine total carbon (not reported). This facilitated the choice of samples for d 13 C and 14 C analyses of dispersed carbonates. Soil samples at site 271007 were also collected to a depth of 0.5 m for 14 C and d 13 C characterisation. The d 13 C of soil total carbon, dispersed carbonates (siderite-d 13 C and d 18 O) and AMS graphite-targets were all analysed by isotope-ratio mass spectrometry (IRMS). 14 C solid samples were analysed by accelerator mass spectrometry (AMS) at the ANSTO STAR accelerator (Fink et al. 2004).
Powdered, randomly oriented drill chip samples were analysed by XRD (X-ray diffraction) with spectra collected from 2 to 80 (2u) on an Xpert Pro diffractometer using Cu Ka radiation with a step increment of 0.03 u and an XCelerator detector. All diffractograms were processed with SIROQUANT (Taylor 1991;Ward et al. 1999) allowing for quantitative proportions of major minerals to be calculated. Samples were not prepared specifically for clay fraction analysis, so a total clay percentage (kaolinite þ illite þ interstratified illite/smectite) is provided together with the rest of the major phases (Table S2).

Groundwater sampling and analytical methods
Monitoring wells from the New South Wales Office of Water (NOW) were accessed for groundwater collection. All wells have PVC casing and are screened at specific depths (Table S1). A total of 50 samples were collected from 18 wells, between 2006 and 2012. Before sampling, all wells were pumped until three well volumes were removed and/or stable parameters were achieved. Purging and sampling was conducted using a submersible Grundfos pump. Specific conductivity (SC), dissolved oxygen (DO), temperature, pH and Eh were monitored in a flow cell with all samples collected from an in-line 0.45 mm, high-volume filter. Details on bottle preparation, field collection and sample preservation can be found in Cend on et al. (2010).

MAJOR AND MINOR IONS
Total alkalinity was determined in the field by direct titration using a digital titrator and standardised H 2 SO 4 acid at 0.1600 N (AE0.0008 N). Selected minor elements were characterised by inductively coupled plasma atomic emission spectroscopy (ICP-AES) (Al, Si, Mn and Fe) or inductively coupled plasma mass spectrometry (ICP-MS) (Sr). Ion chromatography (IC) was used to analyse major anions; major ions were assessed for accuracy by evaluating the charge balance error percentage (CBE%; Table 2).

87
Sr/ 86 Sr ISOTOPES Samples used for trace-element analyses were also used for strontium isotope analyses. The measured Sr concentrations (Table 3) were used to determine the necessary volume of sample to achieve $50 ppb of Sr, taken to dryness and redissolved in HNO 3 and loaded onto pre-conditioned columns packed with Sr-Spec resin. The eluted Sr was loaded onto a single Ta filament with H 2 O and H 3 PO 4 , and oxidised in air. The isotopic ratios were measured on a VG 354 mass-spectrometer at the CSIRO radiogenic isotope facility (North Ryde, Sydney) with a longterm analytical precision of AE0.000014 measured on NBS987 ( 87 Sr/ 86 Sr ¼ 0.710288 AE 14). The 87 Sr/ 86 Sr results were normalised to 86 Sr/ 88 Sr ¼ 0.1194.
Stable water isotope ratios were analysed by IRMS via CO 2 equilibration for d 18 O and an H-Device for d 2 H. Results are reported using the d notation against VSMOW and are accurate to AE1% for d 2 H and AE0.2% for d 18 O. The d 13 C DIC samples were collected in preconditioned gas sealed 12 mL glass vials (Exetainer) after 0.45 mm filtration for samples until 2009 and then 0.22 mm for the rest of samples; no additives were added (Doctor et al. 2008). Samples were maintained in dark refrigerated conditions and analysed shortly after collection. The isotopic signatures of DIC (d 13 C) in the first two sets of samples (2006)(2007) were measured with an Aurora 1030 TIC/TOC interfaced to an IRMS. The remaining samples were measured with a Delta V Advantage mass spectrometer. In-house standards, established by runs with NBS18, NBS19 and LSVEC, were run as samples to allow the results to be properly reported vs VPDB. Results are accurate to AE0.3%.

H AND 14 C DIC ANALYSIS
Samples for both 3 H and 14 C were collected in 1 L Schott glass bottles. 3 H water samples were distilled and electrolytically enriched prior to analysis by liquid scintillation. The 3 H concentrations were expressed in tritium units (TU) with uncertainties ranging from $25% at the quantification limit to <6% for 3 H concentrations above 1.5 TU. The quantification limit varied from 0.4 to 0.13 TU depending on the time of analysis, with analytical improvements reducing quantification limits for more recent samples (e.g. 2011 and 2012). For 14 C analysis, the total DIC was converted to CO 2 by acidifying the samples with H 3 PO 4 and extracting the liberated CO 2 gas using a custom-built extraction line. The CO 2 sample was then heated in a sealed glass tube, containing baked CuO and Ag and Cu wire at 600 C for 2 hto remove any sulfur compounds that may have been liberatedand followed Table 2 Groundwater major physico-chemical parameters, ion concentrations and charge balance error (CBE%).

Geochemical calculations
Calculations necessary to assess electrical neutrality, dissolved element speciation and saturation indices for common minerals phases were undertaken using the WATEQ4F thermodynamic database in the PHREEQC program (Parkhurst & Appelo 1999).
Geochemical models have been applied to account for dilution of 14 C in the aquifer matrix. The code NETPATH (Plummer et al. 1994) has been used to account for mass transfer between initial and final groundwater samples. The initial sample selected to represent non-contaminated shallow groundwater was well 80163, with final samples representing intermediate to deep wells. Details on the input parameters for calculations are discussed below.

Solid samples
Most wells intersected Hawkesbury Sandstone, with deeper wells >100 m most likely reaching the upper Narrabeen Group sandstones, but no apparent changes in colour, lithology or mineralogy were recognised from recovered samples to distinguish the boundary between the two units ( Figure 2). Two sections are recognised in all sampled wells based on mineral distribution, a shallow weathered section devoid of siderite but with its oxidised product, goethite and a deeper section with siderite but without oxidised phases ( Figure 2). The depth separating these two sections is highly variable from 37 m in 271012 to 91 m in 271008 (Table S2). Similar depths separating weathered from non-weathered sandstone have been recorded elsewhere in the Sydney Basin (Pecover 1986).
Sandstone (quartz arenite) comprises the bulk of recovered samples; whether they correspond to sheet or massive facies (Table 1) cannot be readily differentiated from whole-rock fragments. Quartz content in the shallower weathered section is higher with an average of 84 wt% compared with 78 wt% in the lower section. This is consistent with results from nearby Somersby (Krejci 2001) and other parts of the Sydney Basin including the Newnes Plateau (Pecover 1986) and the Southern Highlands (Freed 2005). Shale lenses can be recognised in all wells through sharp increases in the clay content of up to 54 wt% ( Figure 2). The maximum thickness of the intersected clay lenses is estimated at 2 m, with a maximum of 4 individual lenses recognised within wells 271012 and 271009.
Siderite and its oxidation products (e.g. goethite) are localised in the deep and shallower sections, respectively. The goethite contents averaged 1.5 wt% with one value of 41 wt% measured in 271012 that probably corresponds to an Fe-oxide rich layer, generally developed at weathering fronts (Pecover 1986). Siderite content averaged 4 wt% and was generally confined to the deep samples. Siderite in the Sydney Basin has been interpreted mainly as an  nd, not determined. early diagenetic product with some late carbonate cements also present (Al Gahtani 2012;Martel et al. 2012).
K-rich feldspars such as microcline are identified in the deeper parts of wells 271007 and 271012 (average values of 4.5 and 6.3 wt%, respectively), coinciding with the presence of siderite. No microcline was observed in 271008 and 271009. The absence of K-feldspars in 271008/9 along with its absence in the upper part of the other wells suggests weathering. Other minor accessory minerals such as rutile and anatase are observed in all samples with an average value of 0.7%.
The average d 13 C and d 18 O values for siderite recovered from 271007 were d 13 C ¼ À5.1% VPDB and d 18 O ¼ À16.6% VPDB (Table 4). These values are consistent with sedimentation in a braided river system within a transitional depositional environment (Mozley & Wersin 1992). Analysis of 14 C in dispersed siderite samples was carried out in order to rule out the possibility of precipitation of modern siderite, and the results were indistinguishable from 0 pMC (n ¼ 5; Table 4).
The d 13 C samples from total carbon in the soil profile averaged À24.6% (n ¼ 7; Table 4), confirming that the modern soil carbon contribution to recharge water is derived from C3-type vegetation.

Physico-chemical parameters, major and minor ions
All groundwater samples have low salinities with a mean specific conductivity (SC) of 153 mSÁcm À1 . The pH values were slightly acidic and broadly correlated with sample depth, with pH increasing with depth. Exceptions were observed in well 75014 screened within basalt, 271008-2 situated in the central plateau and 271009-3 where preferential flow paths are leading to the change in pH (Table 2). Dissolved O 2 decreased generally with depth with a mean of 1.2 mgÁL À1 . Measurements of Eh relative to the standard hydrogen electrode, while interpreted with caution, showed relatively oxygenated conditions with Eh ranging from 0.67 V to 0.09 V.
Shallow groundwater is generally of the Na-Cl type with a tendency to Na-HCO 3 and eventually Ca-HCO 3 with depth. Exceptions, as observed for pH, correspond to samples screened within basalt (75014), the central plateau (271008-2) and 271009-3 where chemistries are consistent with shallower samples elsewhere in the plateau. Natural nitrate (NO 3 À ) concentrations in shallow wells from non-cleared areas are very low at 0-0.5 mgÁL À1 (e.g. 75015-1) compared with concentrations reaching up to 25.5 mgÁL À1 in shallow wells affected by inputs from agricultural practices (  Quantification limits may vary according to instrument tuning, analytical session and sample matrix but were typically: Al $1.5 mgÁL À1 , Mn $10 mgÁL À1 , Fe $10 mgÁL À1 and Sr $0.5 mgÁL À1 . Sr/ 86 Sr isotope ratios (Table 5). Sr-isotope values range from 0.70956 to 0.73825 indicating that several Sr sources are present in the region. Shallow samples tend to show more radiogenic ratios in non-cleared areas (e.g. 75015-1) while shallow groundwater samples from cleared areas with high NO 3 À concentrations show less radiogenic values (e.g. 271007-1, 75012-1), this being potentially influenced by Sr derived from fertilisers with an average value of 0.70881 (Vit oria et al. 2004). However, marinederived Sr-isotope signatures in shallow groundwater cannot be ruled out, particularly in areas of intense weathering and lower NO 3 À concentrations (e.g. 271008-1). In general, there is a tendency for samples to show increased Sr concentrations and decreased 87 Sr/ 86 Sr ratios with depth. The presence of basalt in the region may also contribute to lower isotopic ratios with depth.

Water stable isotopes (d 2 H and d 18 O)
The stable water isotope signatures in groundwater in the KMMA generally lie close to the coastal Local Meteoric Water Line (LMWL) obtained in the south of Sydney ($100 km from the study site) by Hughes & Crawford (2013), with data from other wells in the Gosford area showing similar characteristics (Table 5). This shows the groundwater is of meteoric origin, but in some cases, generally in shallow wells, the signatures follow an evaporation trend.

Dissolved inorganic carbon ( 13 C DIC )
Dissolved inorganic carbon isotope (d 13 C DIC ) values range from À24.6% in 75013-1 to À5.7% in 271007-2 (Table 5). Shallow groundwater sites have d 13 C DIC values of $À25%, as expected for groundwater interacting with soil organic matter within C3-vegetated areas (Vogel 1993) and from observed total carbon isotopic results of the soil profile (Table 4). Deeper samples show a trend towards more 13 C-enriched values indicating the transition from a CO 2 open system to a closed one with some interaction with dispersed carbonates.

H and 14 C DIC analysis
The 3 H generated during nuclear testing in the 1950-1960s has decayed in Sydney and the Southern hemisphere in general, leaving modern tritium inputs near to natural levels with activities close to $2 TU (Tadros et al. 2014). Values in the KMMA range from 2.3 TU in 75013-1 to below the quantification limit in most deep wells (Table 5). There is a general trend for lower 3 H values with increasing depth; however, the deepest groundwater sample (271009-3) has 3 H values well above the quantification level for two samples in different years (Table 5).
The a 14 C ranges from 116.9 to 1.4 pMC, with most shallow groundwater samples showing a bomb-pulse contribution. The lowest values are found in 75014, coinciding with the basalt aquifer hosted groundwater. Samples repeated over several years tend to show similar results, with the exception of samples in the central area, where 271007-2 and 271008-2 show a tendency to higher a 14 C values while a deeper 271007-3 shows an opposite trend.  (Rosenbaum & Sheppard 1986). IRMS d 13 C analysis of the graphite targets used in the AMS analysis are also included for comparison. The negative a 14 C corresponds to a sample that was at or below lab blank levels.

Sources of major ions
The close proximity to the ocean and low reactivity of local lithologies means that the shallow groundwater would be expected to receive most of its natural solutes from atmospheric deposition via rainfall, aerosols and dry deposition. However, as groundwater moves deeper within the aquifer, initially Na þ -Cl À type water gradually changes to Na þ -HCO 3 À and eventually Ca 2þ -HCO 3 À with depth. The Na þ /Cl À ratios of groundwater are close to those of seawater with a tendency to higher K þ , related to interaction with dispersed K-feldspars in the sandstone (Figure 5a). In the case of 75014, the excess K þ results from groundwater interaction with basaltic minerals that can contain K 2 O concentrations of up to 0.76 wt% in Kulnura basanites (Ian Graham pers. comm.). The absence of shallow thick shale lenses with perched groundwater minimises cation exchange processes close to the surface, while the deeply weathered nature of the sandstone facilitates the quick infiltration of rainfall. The cation exchange capacity (CEC) of noncultivated soils across the study area is quite low and ranges from 2.8 to 43.8 meq kg À1 with changes in depth depending on intersected materials (Hawkins & Haddad 2011) and contrasts with shallow values in weathered shale of 83 meqÁkg À1 observed elsewhere in the Sydney Basin (Temple & Smith 1959). The very low CEC close to surface and the proximity to the coast favours shallow groundwater with ionic ratios close to those of seawater.
Generally, Cl À behaves quite conservatively (Figure 5a, b), but there is a tendency to decrease in concentration with depth; for example, well 75012 changes from an average of 30 mgÁL À1 to 20 mgÁL À1 , from shallow to deep. The higher Cl À concentrations found in the shallower groundwaters are most likely due to land management practises i.e. the use of potash salts as fertiliser (B€ ohlke & Denver 1995), rather than natural variations associated with lower average rainfalls rates. This is also suggested by a weak correlation with NO 3 À concentrations in shallow wells (Figure 5c). The Cl À /Br À vs Cl À ratios show the general conservative behaviour of Cl À and a dispersion of Cl À /Br À that is not associated with sample depth. This can be related to the proximity to taller vegetation along higher ridges enhancing collection of sea-spray aerosols and dry deposition between rainfall events (Whipkey et al. 2000). The halite associated with seawater-derived aerosols and collected by vegetation would be impoverished in Br À , as this element is not a structural component of halite, therefore accounting for its deficiency in groundwater (Cend on et al. 2004). Increases in major ions with depth, particularly Ca 2þ , Mg 2þ and HCO 3 À , follow a linear trend parallel to that expected by dissolution of carbonates (Figure 5d), although with lower divalent cation concentrations, probably owing to an increase in cation exchange with depth as infiltrating groundwater interacts with fresh shale lenses and sandstone. From this plot, two groups are differentiated, roughly separating shallow from deep samples. Iron concentrations can be high and up to 19 mgÁL À1 in the groundwaters (Table 3). In general, Fe concentrations increase with depth because low pH groundwater favours dissolution of Fe-oxides, and also reducing conditions favour Fe to be in solution (Fe 2þ ) (Figure 5e). The difference in concentration with depth suggests variable weathering where Fe is being leached deeper in the aquifer in some areas. Precipitation of Fe-hydroxides in shallow areas owing to water-table fluctuations and related changes in redox conditions may also influence the mobilisation of Fe.

GROUNDWATER ZONES AT KULNURA-MANGROVE MOUNTAIN AQUIFER
Two groundwater zones (shallow and deep) are differentiated in the main aquifers of the study area; depending on the presence and proportions of minerals, inferred degree of weathering, and presence of various hydrogeochemical tracers (major ions, 14 C, 3 H, 87 Sr/ 86 Sr) in the groundwater. The area around well 75014 does not fit into these zones, as this intersects groundwater within one of the local diatremes. The upper groundwater zone extends from the surface to $50 m below in the Mangrove Mountain and Somersby plateaus and down to  $90 m below the Peats Ridge plateau (Figure 2). The upper zone is characterised by modern groundwater with a 14 C values ranging from 116.9 to 72 pMC and measurable tritium ranging from 2.3 to 0.5 TU. The presence of 3 H in general makes this water <70 years old, while 14 C bomb peak modelling suggests groundwater residence times of <20 years in some locations as discussed later. Major ions generally increase with depth, while NO 3 À concentrations follow a reverse pattern decreasing with depth as redox conditions favour denitrification. The majority of the private wells are located within the shallow groundwater zone, which correspondingly has the higher groundwater yields (up to 3 LÁs À1 ). These high yields suggest that the extensive weathering has transformed the upper zone into a semi-porous medium where piston flow is most likely the main form of recharge.
The lower zone generally extends below $50 m and below $90 m in the Peats Ridge area. It is characterised by: a sharp increase in groundwater residence times and 3 H activities below quantification levels, increases in most major ion concentrations consistent with dissolution of dispersed carbonates, very low concentrations or absence of NO 3 À , and low 87 Sr/ 86 Sr ratios likely to be controlled by the dissolution of carbonates. The boundary between the shallow and deep groundwater zones varies across the study site and appears to be related to the sandstone weathering profile and pumping activities in the area.

NETPATH CALCULATIONS
The presence of surface water contaminated with 14 C from post-1950 nuclear weapons testing and fertiliser products precludes the use of most shallow samples as initial groundwater in the calculations to determine the residence times of deeper groundwater (not affected by bomb pulse or contamination). To avoid this problem, a sample from a forested area, with no signs of agricultural inputs and <100 pMC, has been selected as representative of recharge water (80163; Table 6). This sample is therefore used as the 'initial well' in all calculations regardless of connection along a direct flow path. The similarity of the geology and potential processes across the study site ensures that this approach is sound. The initial water is an Na þ -Cl À type with low TDS and contains a 3 H concentration of 1.1 TU and d 13 C (DIC) of À20.4%very similar to that of groundwater incorporating CO 2 from the soil zoneand a 14 C of 84.7 pMC, similar to a composite soil sample of the first metre (Table 4). The model calculation, within the constraints and phases chosen, determines potential reactions that could explain the composition of the final groundwater. An example of the generic reaction used to calculate groundwater ages is given in Equation 1: 80163 þ 0:36 Sid þ 0:13 Cal þ 0:07 Alb ¼ 75012-2-07 À 2:45 CO 2 ðgÞ À 0:12 Goethite where 80163 and 75012-2-07 represent concentrations of initial and final groundwater, respectively. The initial sample is allowed to dissolve dispersed carbonates (Sid ¼ Siderite and Cal ¼ Calcite) and dispersed silicates represented by Alb ¼ Albite; all these 'reactive phases' are unsaturated in all groundwater samples. The numbers in front of the different phases correspond to the mass transfer in mmolÁkg À1 for that specific phase and example. Organic matter is also incorporated and oxidised with the 'product phases' being goethite, which is generally saturated in final samples and identified as a mineral phase by XRD. The CO 2(g) degases along the flow path with shallower samples found to have a higher P CO2 than the final samples. The d 13 C and 14 C of dispersed carbonates and soil materials have been measured and included in calculations, unlike most studies where those sensitive parameters are often assumed (Table 4). Different 14 C correction schemes are then applied to the calculated mass transfer to determine groundwater ages (Table 5). The ages are extrapolated between wells for graphical representation ( Figure 6) with annual variation reflecting the increase in modern groundwater with depth.

TRANSITION FROM SHALLOW TO DEEP GROUNDWATER IN THE KULNURA AND SOMERSBY AREAS
In the Kulnura and Somersby areas, a 14 C values were relatively consistent through all sampled years in both shallow and deep groundwater zones, suggesting a stable system. For example, in the Somersby area (75012) the separation between the shallow and deep zones is situated at about 45-50 m from the surface with the shallow portion showing consistent modern 14 C values (up to 110.8 pMC), 3 H ranging from 2.1 to 1.4 TU, and high NO 3 À concentrations consistent with inputs from the agricultural station nearby. The deep zone at 75012 has very constant a 14 C values ($19 pMC) and an average corrected age of 11.3 ka BP (Table 5, 6). The 3 H activity is always below quantification levels; however low NO 3 À values, $1.8 mgÁL À1 , suggest that nitrate from farming practices has diffused into deeper groundwater at the site. In Kulnura (75015), the limit between both zones is located slightly over the first screened interval (52-55 m) where groundwater a 14 C averages 72 pMC, and 3 H is just above or below quantification limits with groundwater ages ranging from modern to $0.9 ka BP. The deep groundwater samples have an average a 14 C of 25.5 pMC, with similar values for all years and a residence time averaging 8.6 ka BP. Whether the lower mean ages in the north are due to the proximity to the headwaters of the recharge area or due to a slightly shallower position compared with the deep sample in the south (75012) is difficult to ascertain.

TRANSITION FROM SHALLOW TO DEEP GROUNDWATER IN THE CENTRAL AREA
The central part of the study area (W to E) across all three plateaus shows important variations in groundwater residence times controlled by multiple factors. On the Mangrove Mountain plateau, well 75014 has a distinct geological setting. Barron & Lishmund (1998) interpreted the drilled materials at that site as reconstituted sandstones deposited in a maar lake formed during the closing stages of diatreme activity. The materials include K-feldspars, olivine and vesicular basalt fragments, cemented with carbonates (calcite, ankerite and siderite). Carbonate cements reduce the permeability of the aquifer at this location. The oldest residence times were identified in this area with low a 14 C of 1.4-3.2 pMC and corrected ages of 32-26 ka BP. No 87 Sr/ 86 Sr analysis are available from groundwater in this location, but higher TDS compared with other samples in the region (Table 2) and different ionic ratios (Figure 5a, d) clearly show interaction of groundwater with non-Hawkesbury Sandstone materials. Wells closer to the diatreme do not show any influence of mixing between Hawkesbury Sandstone groundwater and that of groundwater interacting in the diatreme, suggesting groundwater from the diatreme system may flow west directly towards discharge points along Mangrove Creek or down into the deeper regional system.
Along the Mangrove Mountain plateau, well 271009 has important local differences; it shows the only depth reversal of ages identified in the study area. Groundwater from the shallower 271009-2 well has an average a 14 C of 44 pMC and average age of 3.85 ka BP while the deeper well (271009-3), screened at $150 m depth, has modern ages with 90 pMC. The 3 H values are similar in both wells despite the 14 C difference, with NO 3 showing large variations in the deepest well (2-21 mgÁL À1 ) and similar results in the shallower well (10.1-11.2 mgÁL À1 ). This suggests a mixture of older and modern waters feeding preferential flow paths to 271009-3. The base of this well is at a similar elevation to the bottom of Mangrove Creek to the west or Popran Creek to the east (Figures 1, 2). Wells T. (personal communication 2012) reported NO 3 À concentrations up to 17.7 mgÁL À1 in Little Mooney Mooney Creek during peak discharge in September-October 2004 with NO 3 À concentrations decreasing sharply to levels of $3.5 mgÁL À1 during low flow conditions. While NO 3 À concentrations in the sampled well may be similar to those in nearby creeks, groundwater levels tend to be lower towards the edge of the plateau with groundwater flowing towards the creeks. The sharp response of water levels (not shown) to rain periods and the high NO 3 À concentrations at the site in the shallower well suggests that this well may be leaking water from the surface and is not representative of the sampled depth.
Upstream of Popran Creek, well 75013-1 shows a 14 C of $100 pMC for shallow groundwater with an increase with depth to an average value of 109 pMC (Table 5) consistent with a 14 C bomb pulse signature. The deeper samples are quite similar to each other with an average a 14 C 4.3 mgÁL À1 in the intermediate piezometer . The Sr-isotope ratios, in groundwater close to the surface, have values consistent with mixtures of Sr from fertiliser and/or marine-derived sources with more radiogenic Sr derived from silicate weathering. Values in the intermediate piezometer are more radiogenic, consistent with a higher influence of sandstone-derived Sr. A trend to less radiogenic ratios is observed in the deepest piezometers and similarly elsewhere in the study area and in other areas of the Sydney Basin (e.g. Lower Blue Mountains, unpublished results) representing an increasing interaction of groundwater with Permian-Triassic marine-derived Sr preserved in the dispersed carbonates.
In the Peats Ridge plateau, considerable changes in groundwater residence times are observed. Shallow samples show high 14 C bomb pulse signatures, indicating modern recharged groundwater (Table 5; 271008-1), while deeper groundwater shows an increasingly modern 14 C signature with time, instead of lower a 14 C values, as observed in other wells. Values evolved from 36.1 pMC and 5.2 ka BP in 2007, similar to those in 75013-3 at a similar depth, to modern values of 103 pMC in 2010 with the latest sample in 2012 failing to graphitise, probably owing to the high CO 2 generally linked in the study area with modern groundwater. The 3 H have also evolved from values below quantification limit in 2007 and 2008 to values of $1.1 TU slightly lower than those found in shallower groundwater at the site. Sharp variations in the SWL through the year indicate that groundwater at this location is affected by nearby pumping activities. Remarkably, groundwater levels rebound once pumping ceases, showing the dynamic behaviour of the groundwater system near this bore and suggesting induced recharge from shallow upstream areas. The mineralogy of solid samples in 271008 suggests deep weathering to depths of 90 m, which allows for increased groundwater storage. Pumping activities are leading to NO 3 À contamination at depth with a progressive yearly increase in concentration (2007)(2008)(2009)(2010)(2011)(2012) observed from below detection limit to 6.1 mgÁL À1 . The NO 3 À concentrations in the shallow areas around 271008 are low ($3 mgÁL À1 ) suggesting NO 3 À is flowing to the well from shallow groundwater in areas further afield. Other major ions have decreased in concentration (Ca 2þ , HCO 3 À ) with groundwater becoming more acidic over the years at depth and increasing the concentration of metals such as Al. Sr-isotopes are also consistent with the recharge of modern groundwater in depth with a tendency to less radiogenic ratios with time (Figure 5f). Similar trends to those observed at 271008 can also be observed at 271007 although likely in earlier stages of chemical development. Whether hydrogeochemical tracers at 271007 are being affected by pumping further afield or responding to local pumping cannot be discerned. Shallow samples have groundwater with a 14 C clearly affected by bomb pulse inputs (116.9 pMC). Deeper samples show an increase in a 14 C over the sampled years with 3 H also increasing to levels of 2 TU. Srisotope ratios close to the surface are low (0.710915). This, combined with the highest NO 3 À concentrations found in this study (up to 25.5 mgÁL À1 ), suggests that Sr derived from fertilisers may have overwhelmed more radiogenic signatures typical of local groundwater in contact with sandstones containing dispersed K-feldspars and generally lower total Sr concentrations. Deeper groundwater samples at 271007-3 show a tendency towards lower a 14 C (48.2 in 2007C (48.2 in to 38.2 pMC in 2009) and residence times from 2.5 to 5.2 ka BP that may suggest the system was stressed by regional pumping, especially during the drought, as the SWL has been steadily recovering since 2007. The temporary but significant depressurisation at depth may have resulted in older stagnant water below the plateau being mobilised towards the production bores past this sampling point. Another contributor to old signatures may involve diffusive exchanges of tracers between young water in transmissive fractures and much older stagnant water in adjacent matrix pores. This is most likely where fractured rocks occur at depth, given sufficient periods for interaction. Extended periods between pumping would result in some young age tracers diffusing from the fractures into the older matrix water, and vice versa. Such diffusion can lead to the sampled water, transmitted through the fractures, taking on a mixed age signature.
Alkhatib & Merrick (2006) used particle tracking to calculate the time necessary for water to move from the top layer of their model (surface) to the lower layer at approximately sea-level, broadly coinciding with the base of the creeks draining the plateaus and with the deeper samples analysed in this study. They found that residence times in the order of 2.5 ka were required to reach the lower layers of their model and coincides with results obtained from 14 C in the Central Mangrove Mountain area (271007-3-07); however model calculations may not be able to account for the inter annual variation observed, probably owing to pumping and movement of older water as observed in 14 C residence times. Additionally, to the north and south, the 14 C results presented in this study are generally older than predicted by the particle-tracking mode, and while it is not clear if particle tracking results can be extrapolated to those areas, the transition to a fractured-controlled flow may account for some of the discrepancy.

BOMB PULSE AND MODERN GROUNDWATER AGES
Bomb pulse 'age calibrations' are generally applied to closed systems with respect to carbon (Hua et al. 2013). In a dynamic and 'open' system such as that which is expected in this shallow groundwater system, it can only be used as a rough proxy. Soils in the study area are thin, and 14 C analysis of soil in the top 35 cm, as analysed at 271007, shows an average of $106 pMC similar to modern rainwater. The quartz-rich nature of the aquifer matrix ensures minimum dilution of the dissolved 14 C, while the porous nature of the shallow aquifer ensures a fast recharge. Chloride mass-balance estimations of recharge, using the saturated zone method (Wood & Sanford 1995) and a conservative value for atmospheric Cl input of 1.69 mgÁL À1 obtained at similar altitudes from rainfall at Saddleback Mountain (south of the study area; Golab 1998), provide an average recharge value of 108 mmÁy À1 or about 9% of rainfall with higher values up to 14% in the Peats Ridge area, consistent with high recharge.
Seven shallow wells show bomb pulse signatures (Table 5; Figure 7) whereas several other groundwater samples have values only slightly over 100 pMC, consistent with modern rainfall values. The highest a 14 C value was measured in 2007 in 271007-1 (116.9 pMC) while the sample collected in the same well the following year was 112 pMC. The southern hemisphere peak for a 14 C reached close to 170 pMC, while the 3 H was $60 TU for Melbourne, both in 1965. If we assume water was recharged in the 1960s, the system would have behaved in a closed manner, and 3 H should have higher activities $3 TU, which is not consistent with chloride recharge estimates. Furthermore, the early 1960s experienced below-average rainfall (Gosford Station); however, in 1990, the region received record maximum rainfall with important associated recharge expected (Figure 7). Tritium concentrations for 1990 rainfall weighted samples averaged 4.9 TU at Campbelltown ($50 km SW of Sydney) (Tadros et al. 2014), consistent with decay-corrected values found in shallow samples (271007-1).
Considering the multiple points of evidence as described above, we argue that 14 C indicates that residence times for shallow groundwater, at least in 271007, are in the order of ca 17 y. Even assuming a uniform porosity of 15%, the range of estimated vertical hydraulic conductivities, K v , based on apparent vertical velocities for the entire system, is wide (Table S3). For example, in the upper 50 m of 75013, K v values are around 6 Â 10 À8 mÁs À1 but are several orders of magnitude less conductive at depth. Indeed, most intervals below about 250 m AHD across the region showed K v values below about 5 Â 10 À10 mÁs À1 , whereas most intervals above that elevation were above 10 À8 mÁs À1 . Exceptions were the poorly conductive shallow well 75104, above 260 m AHD in the diatreme, and the highly conductive upper 40 m of 75012, despite lying below 250 m AHD towards the edge of the Somersby Plateau. Most of the low K v values may be one or two orders of magnitude lower than corresponding horizontal hydraulic conductivity, K h , as for many of the Sydney Basin formations. Where weathering and leaching have significantly modified the porosities and mineral orientations, vertical and horizontal hydraulic conductivities may be similar.

CHEMICAL CHANGES INDUCED BY PUMPING
Land and groundwater usage has the potential to influence the groundwater chemistry of an aquifer. The addition of fertiliser-derived NO 3 À (discussed below) is a clear example of this; however this was not the only chemical impact on groundwater observed. On a regional scale, groundwater recharges quickly minimising potential evaporation during infiltration, as supported by the close proximity of most samples to the local meteoric line (Figure 7b). However, in some sections of the aquifer, irrigation practices have locally changed this dynamic with all evaporated samples (e.g. 75013-2) located in irrigated areas where generally open wells drawing groundwater from intermediate to deeper parts of the system which is pumped to the surface and stored in ponds prior being used for irrigation. Shallower samples in 75013-1 do not show the same enrichment compared with the deeper 75013-2 samples suggesting recharge during wet conditions with limited irrigation input. The correlation between NO 3 À concentrations and H 2 O-d 2 H (in single locations and regionally) further supports the link between increased evaporation and agricultural practices (Figure 8a). The minimal buffering capacity of the quartzose sandstone aquifer, at least in its upper zone where dispersed carbonates have been long dissolved, means that shallow groundwater generally has a low pH. A limited set of results from 1998 (Jiwan & Williams 1998) shows a higher pH for all samples compared with the same wells analysed for this work. However, it is in the central area around 271008 where pH changes are most evident and probably linked to pumping activities. During 2007, groundwater pH was similar to that expected for samples at similar depths ( Figure 8b) with consistent groundwater residence times; however successive samples show a shift to lower pHs similar to those found in much shallower samples, as well as modern groundwater ages. Groundwater extraction is causing an inflow of modern waters at depth with associated acidification of groundwater.
An important consequence of acidification is the capacity to mobilise trace metals; of particular interest is aluminium that has been linked to enhanced risks of cognitive decline, dementia and Alzheimer's disease for subjects with a high daily intake of aluminium from drinking-water (!3.7 mMÁday À1 ; Rondeau et al. 2009). Shallow samples in the Mangrove Mountain area and some of the deeper samples with Al concentrations of $3.45 mM would fit in that risk for average drinkingwater intakes (Figure 8c). The movement of low pH shallow groundwater is causing an increase in Al concentrations, particularly in the central area around Peats Ridge; this may be affecting groundwater for local consumption or that recovered in bottling plants.

NITRATE AND OTHER POTENTIAL CONTAMINANTS
Diffuse regional contamination by nitrates in the Mangrove Mountain area related to farming activities is a problem influencing groundwater chemistry in this system (Bish et al. 1994). Typical applications for citrus are in the order of 0.6 TÁha À1 of ammonium nitrate (NSW Agriculture 2003) with other sources of nutrients like poultry litter also used in the area. Only 22% of the 78 open production wells analysed by Bish et al. (1994), had NO 3 concentrations above 50 mgÁL À1 . General awareness and land-management advances have improved the situation, as none of the 50 groundwater samples analysed in this study exceeded Australian trigger values of 50 mgÁL À1 (as NO 3 À ), with the highest value (25.5 mgÁL À1 ) identified at 271007-1 (Table 2). All other high values were concentrated in shallow wells with the exception of 271009-3, where well construction may be affecting results. Postma et al. (1991) identified that for a substantial nitrate reduction in the aquifers to take place, an adequate reduction potential within sediments was necessary with phases like organic matter, pyrite and Fe 2þsilicates as major electron donors; no pyrite or Fesilicates have been identified. Average dissolved organic carbon in water in the KMMA is 0.54 mgÁL À1 (n ¼ 13), suggesting a limited interaction with organic matter. The absence of these phases could explain why NO 3 À concentrations are high in deep zones particularly in the central area (271008). Whether the peak of the nitrate plume has been reached at depth (271008-2) cannot be assessed, but concentrations have increased steadily from below the detection limit in 2007 to 6.1 mgÁL À1 in 2012. The presence of NO 3 À provides another important indicator of modern groundwater recharge. Vit oria et al. (2004) reported 87 Sr/ 86 Sr ratios for 27 widely available commercial fertilisers covering straight (one main nutrient), to compound (at least two main nutrients) and other fertilisers (secondary nutrients or trace elements) obtaining a median ratio value of 0.70881 to lie below the modern seawater Sr ratio. Strontium concentrations in these fertilisers vary, however, and Senesi et al. (1983) reported a mean value of 28 mg kg À1 for ammonium nitrate with other studies providing ranges from 0 to 2000 mg kg À1 , depending on fertiliser origin (Otero et al. 2005). The addition of fertilisers to nutrient-poor soils with low CEC may be enough to overwhelm an originally marine-derived Sr signature (0.7092) or a more radiogenic ratio typical of groundwater interacting with K-feldspar type minerals (e.g. 0.738246 at 75015-1). Marine-derived Sr would have a similar isotopic ratio to that reported for fertilisers, and available results, while limited (Table 5; Figure 5f), suggest that despite general NO 3 À concentration decreasing with depth owing to dilution and natural denitrification processes (Korom 1992), 87 Sr/ 86 Sr can help to trace the original impact of fertilisers, even if NO 3 À has been totally reduced.

SUMMARY AND CONCLUSIONS
This study combines hydrogeochemistry, isotopic tracers in groundwater and whole-rock samples with a multi-year detailed sampling across the dissected elevated sandstone plateaus of Kulnura-Mangrove Mountain (NSW, Australia). The area relies on groundwater extraction for local drinking-water, agricultural and industrial activities with natural groundwater flow also supporting base flow to the deeply incised creeks surrounding it. The KMMA is mostly hosted in its upper part by the Hawkesbury Sandstone, where intense and deep sandstone weathering profiles have resulted in enhanced groundwater storage. Weathering reactions favoured by the local geological setting have transformed the original Hawkesbury Sandstone quartz arenite into a semi-solid or friable sandstone with variable weathering depths where most of the original carbonate cements have been leached, resulting in higher porosity and permeability. XRD analyses show an upper zone down to $50 m and even 90 m in some areas where all carbonates and probably feldspars have been dissolved, and the derived products goethite and kaolinite have formed. With depth, carbonates (mostly siderite) are present, representing fresher or less-weathered sandstone. Isotopic analysis of dispersed carbonates shows values consistent with their depositional environment and devoid of 14 C. Groundwater hydrogeochemistry shows that most solutes are derived from atmospheric deposition via rainfall, aerosol or dry deposition consistent with water stable isotopes and values close to the local meteoric line. Shallow groundwater is of the Na-Cl type with low pH, very low dissolved solids (SC % 115 mSÁcm À1 ), oxidising conditions and showing the impact of agricultural activities in cleared areas with high NO 3 À concentrations and localised evidence of irrigation return waters indicated by evaporated stable isotope signatures. The low buffering capacity of the shallow quartz arenite aquifer matrix favours the lowering of the groundwater pH by soil-derived CO 2 that then has the capacity to dissolve Al in some shallow groundwaters. 3 H and 14 C show a modern groundwater with conservative Cl mass balance suggesting average recharge values of 108 mmÁy À1 or up to 14% of rainfall in some areas. Bomb pulse signatures in some of the groundwaters, when compared with average cumulative rainfall, while taken with caution, suggest some shallow groundwater recharged in the early 1990s.
Results suggest that with depth, groundwater interacts with fresher sandstone containing dispersed carbonates (siderite) and K-feldspars (microcline). Groundwater tends to a Ca 2þ -HCO 3 À type with higher pH and dissolved solids, and is still oxidising. Sr-isotope ratios show a tendency to less radiogenic values with depth, consistent with dissolution of marine influenced carbonates of Permian-Triassic origin. In the fresher sandstone, the aquifer permeability is reduced, with fracture flow likely to play a more important role with depth and higher groundwater mean residence times reflecting local heterogeneities (the presence of shale lenses, massive or sheet facies). In general, in the north (Kulnura), deep corrected residence times are 8.6 ka BP with higher values in the south (Somersby) 11.3 ka BP. These are consistent with the general southerly flow of groundwater in the aquifer. North and south sections of the study area show similar groundwater ages without relevant inter-annual variations.
The central part of the KMMA around Peats Ridge, while having the deepest weathering profile in the area and therefore enhanced groundwater storage, also shows evidence of the inflow of modern groundwater with depth. While standing water levels seem to recover after pumping, groundwater in 2007 had a corrected mean average age of 5.2 ka BP (36.1 pMC), while at the same location groundwater was found to be younger with subsequent samplings. Interestingly, this sample was found to be modern in 2010 with 103.5 pMC. All other tracers were consistent with the presence of modern groundwater with detected 3 H activity, lower pH, elevated NO 3 À concentrations and increased Al concentrations. The inflow of modern groundwater was also detected further east in the central area, while the western side did not show changes in depth with residence times of 5.3 ka BP similar to what was observed in the central area in 2007, prior to the inflow of modern groundwater.