Identifying ﬂood recharge and inter-aquifer connectivity using multiple isotopes in subtropical Australia

. An understanding of hydrological processes is vital for the sustainable management of groundwater resources, especially in areas where an aquifer interacts with surface water systems or where aquifer interconnectivity occurs. This is particularly important in areas that are subjected to frequent drought/ﬂood cycles, such as the Cressbrook Creek catchment in Southeast Queensland, Australia. In order to understand the hydrological response to ﬂooding and to identify inter-aquifer connectivity, multiple isotopes ( δ 2 H, δ 18 O, 87 Sr / 86 Sr, 3 H and 14 C) were used in this study in conjunction with a comprehensive hydrochemical assessment, based on data collected 6 months after severe ﬂooding in 2011. The relatively depleted stable isotope signatures

−5.34 to −5.13 ‰ VSMOW) were evident in surface water samples (δ 2 H: −25.2 to −23.2 ‰, δ 18 O: −3.9 to −3.6 ‰ VSMOW), indicating that these extreme events were a major source of recharge to the dam in the catchment headwaters. Furthermore, stable isotopes confirmed that the flood generated significant recharge to the alluvium in the lower part of the catchment, particularly in areas where interactions between surface waters and groundwater were identified and where diffuse aquifer recharge is normally limited by a thick (approximately 10 m) and relatively impermeable unsaturated zone. However, in the upper parts of the catchment where recharge generally occurs more rapidly due to the dominance of coarse-grained sediments in the unsaturated zone, the stable isotope signature of groundwater resembles the longer-term average rainfall values (δ 2 H: −12.6, δ 18 O: −3.4 ‰ VSMOW), highlighting that recharge was sourced from smaller rainfall events that occurred subsequent to the flooding. Interactions between the bedrock aquifers and the alluvium were identified at several sites in the lower part of the catchment based on 87 Sr / 86 Sr ratios; this was also supported by the hydrochemical assessment, which included the modelling of evaporation trends and saturation indices. The integrated approach used in this study facilitated the identification of hydrological processes over different spatial and temporal scales, and the method can be applied to other complex geological settings with variable climatic conditions. A. C. King: Identifying flood recharge and inter-aquifer connectivity in subtropical Australia ferent groundwater sources and the spatial variability of this recharge (Hrachowitz et al., 2011;Dogramaci et al., 2012).
While it is generally recognised that recharge is variable over time, the influence of episodic climatic events such as flooding are not very well understood. This is particularly the case in alluvial aquifers where total recharge is often dominated by flood-related influxes (e.g. Workman and Serrano, 1999). In these alluvial systems, recharge rates are commonly elevated during floods, as a result of (1) the enhanced permeability of the creek bed during the flood, due to scouring of the clogging layer by high velocity flows (e.g. Cendón et al., 2010;Simpson and Meixner, 2012); (2) enlarged pathway between surface-and groundwater, due to the increased width of the creek and the interface between groundwater and the creek across which interaction can occur (e.g. Lange, 2005); and (3) the increased head gradient between the creek and the stream (e.g. Rushton and Tomlinson, 1979). Owing to this reliance on infrequent flooding and large rainfall events, alluvial aquifers are likely to be severely impacted by the predicted changes in climatic patterns, such as the projected increased frequency and severity of droughts and floods (Parry et al., 2007). This forecasted climate change will impact on river flows (Arnell and Gosling, 2013) and groundwater recharge processes (Green et al., 2011;Barron et al., 2012;. This is particularly relevant for alluvial systems which are connected to ephemeral or intermittent streams, as interactions between these streams and the alluvial aquifers are highly dependent on antecedent rainfalls (Hughes et al., 2011).
The study area is a small subtropical catchment in Southeast Queensland, Australia, which was subject to severe climate extremes in recent years, including an extended drought from the late 1990s through to approximately 2009, followed by heavy rains, which culminated in a 1 % annual exceedance probability (AEP) flood in January 2011 (Babister and Retallick, 2011). This event provided a unique opportunity to study groundwater recharge processes that result from episodic flooding.
Seepage to the alluvium from the underlying bedrock aquifers is potentially an important source of recharge for the alluvium, but this process has not been verified. The influx of poor-quality groundwater, which is often associated with bedrock aquifers in the study area, may negatively impact on the water quality of the alluvial aquifer. Therefore, it is important to identify and monitor areas where bedrock seepage occurs.
The objective of this study is to demonstrate how multiple environmental isotopes (δ 2 H, δ 18 O, 87 Sr / 86 Sr, 3 H and 14 C) in combination with a comprehensive hydrochemical assessment can be applied to (1) assess the significance of floods as a major recharge source; (2) identify recharge processes and connectivity between surface water and groundwater; and (3) identify areas where the alluvium is recharged by the underlying highly diverse bedrock (inter-aquifer connectivity). Multiple isotopes are increasingly being used to identify inter-aquifer connectivity (e.g. Dogramaci and Herczeg, 2002;Raiber et al., 2009, Cartwright et al., 2010aCostelloe et al., 2012;Baudron et al., 2014); nevertheless, studies of this kind are still challenging due to the complexity of the hydrochemical interactions that result from inter-aquifer groundwater flows.
Many studies have used surface-and groundwater compositions (i.e. isotopes, and major and minor ions) to report on the connection between streams and alluvial groundwater (e.g. Soulsby, 2007;Barrett et al., 1999;Kirchner et al., 2010;Mandal et al., 2011;Morgenstern et al., 2010;Siwek et al., 2011;Négrel and Petelet-Giraud, 2005). However, studies that use isotopes and hydrochemistry to assess the connectivity between alluvial aquifers and intermittent or ephemeral streams (e.g. Kumar et al., 2009;Vanderzalm et al., 2011), or report specifically on the effects of episodic groundwater recharge from flooding (e.g. Cartwright et al., 2010b;Cendón et al., 2010;Simpson et al., 2013) are less common. This study uses groundwater stable isotopes together with a detailed assessment of δ 2 H and δ 18 O in rainfall to assess episodic recharge. Rainfall isotope time series data are commonly used to assess long-term trends in groundwater recharge (e.g. Zhu et al., 2007;Praamsma et al., 2009); however, they are rarely applied to assess event recharge of shallow aquifers (e.g. Scholl et al., 2004;Gleeson et al., 2009). The value of considering time series data of rainfall stable isotopes in hydrogeological investigations is clearly demonstrated by this study, and the outcomes will be important for the management of the alluvial groundwater resources of the study area and for understanding flood-related processes in similar alluvial settings.

Hydrogeological setting
The Cressbrook Creek catchment covers an area of approximately 200 km 2 in Southeast Queensland, Australia. The area considered for this study extends from the Cressbrook Dam in the headwaters to the confluence with the Brisbane River in the northeast; it excludes the area up-gradient of Cressbrook Dam, which is a drinking water supply dam for the Toowoomba City Council (Fig. 1). The topographically elevated areas in the southwest of the catchment (ranging from 220 to 520 m Australian Height Datum, AHD) are forested and mostly undeveloped, whereas alluvial plains along the drainage system host rich farm land (> 90 % of the total alluvium by area), particularly in the lower part of the catchment to the northeast (approximately 70-150 m AHD). In this part of the catchment, irrigators use up to 3 GL of alluvial groundwater annually (DNRM, 2012), but groundwater abstraction is often restricted due to low groundwater levels. With the construction of Cressbrook Dam in 1983, flow in Cressbrook Creek was further reduced, resulting in lower groundwater levels. While water was initially released from the dam to recharge the alluvium, releases were controver- sially phased out in the late 1990s due to drought-induced water shortages. In this study, the catchment has been arbitrarily divided into four regions for ease of discussion: the catchment headwaters, the upper catchment, the mid catchment and the lower catchment (Fig. 2).
This climatic variability has been particularly evident in recent years, when below average rainfall from 2000 to 2009 resulted in very low creek flow, especially from mid-2006 until early 2008 when flow in the creek ceased completely (Fig. 4b). Due to that extended drought, water levels at Cressbrook Dam in the headwaters of the catchment (Fig. 2) did not reach the overflow in the period between 1999 and early 2011, and there was no flow from the dam to the creek. Despite the lack of outflow from the dam, intermittent flow was recorded in Cressbrook Creek during this period of time (Fig. 4b), indicating that the creek was recharged by both overland and groundwater contributions along its course. The period of drought was then followed by 2 wet years (2010 and 2011), culminating in significant flooding in January 2011 (Fig. 4b), approximately 5 months prior to the sampling conducted during this study. As a result of this flooding, Cressbrook Dam reached the overflow and discharged to Cressbrook Creek until 24 June 2011, with peak flows of approximately 330 m 3 s −1 . During the surface water sampling campaign (7-8 June 2011), approximately 0.5 m 3 s −1 was discharging from Cressbrook Dam (Toowoomba Regional Council, 2012) and Cressbrook Creek was flowing at approximately 0.7 m 3 s −1 at CC3 (  brook Creek was probably derived from the dam during this period. Groundwater hydrographs show that during the peak of the drought in 2008, groundwater levels had dropped to approximately 4-5 m below the base of the creek in the lower catchment. Additionally, the groundwater gradient in the lower catchment indicated that the creek was losing during this drought period. However, groundwater levels recovered following the flooding and heavy rain in 2010 to 2011. Subsequent to the flood, the groundwater gradient reversed and Cressbrook Creek became a gaining stream in the mid to lower catchment (Fig. 3), suggesting that groundwater gradients between the alluvial aquifer and stream are dynamic and dependent on the antecedent rainfall conditions. However, it is apparent that the alluvium receives substantial recharge from Cressbrook Creek in the mid to lower catchment (King et al., 2014).

Bedrock
The alluvial aquifer system of Cressbrook Creek overlies bedrock of variable geology, with volcanic rocks, metamorphic rocks and granodiorite prominent in the upper part of the catchment (Figs. 2 and 3). Basaltic rocks are particularly prominent in the upper catchment, whereas the bedrock in the mid to lower catchment is composed mainly of the Mesozoic sedimentary rocks of the Esk Formation. Primary porosities of these bedrock units are generally low, but permeabilities are enhanced in some regions by weathering of granodiorites and fracturing in other rocks (GSQ and IWSC, 1973). The screened (slotted) section is 3 m long at B92 and B158. The values in parentheses in column 2 represent the length of the screened section that is encompassed by each geological material.
The Esk Formation underlies many of the alluvial sampling sites in the mid to lower catchment (Fig. 3), and has a broad range of sedimentary strata and grain sizes (Cranfield et al., 2001). Geological borehole logs (DNRM, 2012) confirm that this formation is very heterogeneous, with clayey sandstones, feldspathic sandstones, shale and basalt, all recorded at shallow depths within the mid to lower catchment.

Alluvium
The alluvial system at Cressbrook Creek is characterised by fining-upwards sequences, which typically consist of basal sands and gravels, overlain by silts and clays. Minor carbonate veins have been identified within granodiorites (Zahawi, 1972). However, their contribution to the alluvium, if any, has not been detected in X-ray diffraction (XRD) analyses of sediments collected from Lake Wivenhoe, which is located downstream of the confluence with the Cressbrook Creek and the Brisbane River ( Fig. 1). In addition, no carbonate was detected in the weathered granodiorite profile (Douglas et al., 2007) as any potential carbonate particles are likely to dissolve. This apparent lack of carbonates implies that radiocarbon dating of alluvial groundwaters is unlikely to be significantly affected by interactions with carbonate minerals. King et al. (2014) describe this complex, multi-layered alluvial system as a two-layer system based on sediment grain size assessment. The basal coarse-grained layer consists mostly of sands and gravels, whereas the upper low permeability layer is primarily composed of fine-grained sediments such as silts and clays. This fining upwards sequence is characteristic of many alluvial systems in eastern Australia (e.g. Cendón et al., 2010;Cox et al., 2013), largely due to diminishing surface water flows in the late Quaternary (Knighton and Nanson, 2000;Maroulis et al., 2007;Nanson et al., 2008). The thickness of the low-permeability layer increases with distance downstream, whereas the thickness of the basal high-permeability layer decreases down-gradient; these variations suggest that there is probably less recharge in the lower parts of the catchment compared to the upper parts.

Water sampling and analytical methods
Surface-and groundwater samples were collected in June and September 2011 from eight surface water sites, 18 bores screened in the alluvial aquifer and eight bedrock bores. In addition, two samples were collected from bores where the screened intervals (slotted section of casing) encompass both the lower 1-2 m of the alluvium and the top 1-2 m of the bedrock (B92 and B158); these sites are categorised as "bedrock sites" (Fig. 3). Alluvial boreholes are less than 20 m deep and they usually have a 3 m long screened section at the base of the alluvium, whereas bedrock boreholes are generally deep, except for three shallow bores screened in the Esk Formation (Table 1).
Prior to sampling, three well volumes were pumped from the boreholes and the specific (electrical) conductance (SC), temperature, redox potential (Eh) and pH were monitored using a flow cell to ensure that these parameters had stabilised prior to sampling. Field measurements were taken with a TPS 90 FL field meter, which was calibrated in accordance with the manufacturer's specifications prior to use.

Major and minor ions
Samples for major and minor cations (Na, K, Ca, Mg, Fe, Mn, Al and Sr) were collected in acid-cleaned 125 mL HDPE (high-density polyethylene) bottles and acidified to approximately pH 2 using HNO 3 . Cations were analysed at Queensland University of Technology (QUT) by inductively coupled plasma optical emission spectroscopy (ICP-OES). Samples for major anion analyses (Cl, NO 3 , SO 4 and HCO 3 ) were collected in pre-rinsed 250 mL HDPE bottles, with no further treatment until analysis, which was performed at QUT using an automated discrete analyser (Seal AQ2), ion chro-matography (Dionex ICS-2100) and by manual titration for alkalinity.

Isotopes
Stable isotopes (δ 2 H and δ 18 O) of groundwater and surface water samples collected during this study were analysed using a Los Gatos Liquid Water Isotope Analyser at the University of New South Wales (after Lis et al., 2008). The δ 13 C of dissolved inorganic carbon (DIC) was analysed at GNS Science (New Zealand). Strontium isotopes were analysed using multi-collector inductively coupled plasma mass spectrometry (MC-ICP-MS) at the University of Melbourne following the methods described by Hagedorn et al. (2011). The internal precision (2se) and external precision (2sd) for the MC-ICP-MS procedure is ∼ ±0.000020 and ±0.000040, respectively. Tritium and radiocarbon were analysed at the Australian Nuclear Science and Technology Organisation (ANSTO). For 14 C analysis, the total DIC was converted to CO 2 using a custom-built extraction line. The CO 2 sample was then graphitised, and graphite targets were analysed by AMS at ANSTO's STAR accelerator following procedures of Fink et al. (2004). Conventional radiocarbon ages were reported as percentage Modern Carbon (pMC) with 1σ errors of less than 0.37 pMC (Stuiver and Polach, 1977). Samples for 3 H analysis were distilled and electrolytically enriched, and subsequently analysed using a liquid scintillation counter. Results are reported in tritium units (TU) with an uncertainty of ±0.04-0.08 TU and quantification limits of 0.13 TU.
Rainfall from Brisbane Airport was collected as a monthly composite of daily rain gauge samples, following the technical procedure recommended for GNIP sampling (http://www-naweb.iaea.org/napc/ih/documents/userupdate/ sampling.pdf). Samples from June to October 2010 were analysed by Isotope Ratio Mass Spectrometry at the CSIRO Land and Water Isotope Lab (Adelaide) (reported accuracy of ±1.0 and ±0.15 ‰ for δ 2 H and δ 18 O, respectively) or Alberta Innovates Technology Futures Isotope Hydrology and Geochemistry Lab (reported accuracy of ±1.0, ±0.2 ‰ for δ 2 H and δ 18 O, respectively). Samples from November 2010 to June 2011 were analysed at the ANSTO Institute for Environmental Research using a Cavity Ring-Down Spectroscopy method on a Picarro L2120-I Water Analyser (reported accuracy of ±1.0 and ±0.2 ‰ for δ 2 H and δ 18 O, respectively).

Geochemical calculations
Evaporation curves and saturation state calculations were performed using PHREEQC (Parkhurst and Appelo, 1999). Evaporation curves were calculated under the assumption that calcite, dolomite and gypsum precipitate when they reach saturation and are not re-dissolved. Mineral stability diagrams were calculated after Drever (1997), using groundwater analyses collected as part of this study.

Aquifer testing methodology
Rising/falling head tests were conducted in order to estimate the hydraulic conductivity of the aquifer at each site. This information was consequently used to estimate recharge, as described later in the study. Rising head tests were conducted using a bailer, or, for the more permeable sites, two bailers joined end-to-end. Falling head tests were performed by pouring water into the bore, and the response was measured using a pressure transducer that recorded the water level at 1 s intervals. The hydraulic conductivity was then calculated using the Hvorslev method (Hvorslev, 1951).

Aquifer testing
Hydraulic conductivity generally decreases with distance downstream, based on falling/rising head tests that were conducted as part of this study (Table 2). Also, the more permeable alluvium is generally located close to Cressbrook Creek.

Hydrochemistry
Surface waters are generally fresh (SC < 850 µS cm −1 ; Ta  to brackish (SC 369-5930 µS cm −1 ) with no clear dominant major cations and low SO 4 concentrations, with SO 4 / Cl molar ratio ranges from 0.001 to 0.21. The Cl / HCO 3 molar ratio ranges from 2.9 to 33.9, with ratios increasing with salinity. The hydrochemistry of the bedrock groundwaters is highly variable, although the Na / Cl ratio is generally higher than in alluvial waters (Fig. 6).

Mineralogy and geochemical interactions with groundwater
To assess the interaction of groundwater with minerals in the soil zone and the aquifer matrix, groundwater hydrochemical data were incorporated into silicate stability diagrams ( Fig. 7) to determine the relative stability of common silicate minerals in equilibrium with groundwater collected from major bedrock aquifers (Esk Formation and the Eskdale Igneous Complex) and the alluvium (Fig. 7). The silicate stability diagrams show that kaolinite is usually in equilibrium with groundwaters from the Cressbrook Creek catchment, except for Ca-rich minerals, which are generally in equilibrium with smectite.

Stable isotopes (δ 2 H and δ 18 O)
Isotopic signatures for groundwater and surface water (Table 4) are compared to rainfall data collected from Brisbane Airport and Toowoomba (Fig. 8) between May 2008 and May 2010 (Crosbie et al., 2012), and new data collected by ANSTO between June 2010 and June 2011 (Table 5). Rainfall collected from the Brisbane Airport ( Fig. 1), located approximately 60 km east of the study site, is isotopically similar to rainfall collected from Toowoomba, which is located approximately 20 km to the southwest ( Fig. 8a; Crosbie et al., 2012). This suggests that there is limited spatial variation in the study region, and that data from Brisbane and Toowoomba are representative of the Cressbrook Creek catchment. The Brisbane Meteoric Water Line (MWL) has a slope of 7.9 , which is close to the global average of 8.2 (Rozanski et al., 1993). However, the deuterium excess (d) of 13.1 ‰ is higher than the global average of about 10 ‰, as observed in other coastal eastern Australian sites (Cendón et al., 2014), probably due to the influence of convective rainfall (Liu et al., 2010). During the 12 months prior to the June 2011 sampling campaign, rainfall stable isotope signatures were depleted compared to previous rainfall events, particularly during, and immediately prior to, the flooding in January 2011. Rainfall from December 2010 and January 2011 (316 and 424 mm respectively; BOM, 2012) was particularly depleted in δ 2 H (−30.2 and −27.8, respectively) and δ 18 O (−5.34 and −5.13, respectively; Table 5) compared to the weighted average for rainfall, which was −3.4 and −12.7 for δ 2 H and δ 18 O respectively (Crosbie et al. 2012). This confirms observations by Hughes and Crawford (2013), who also noted that high-precipitation rainfall events associated with east coast pressure systems in Australia can be significantly depleted. The slope of the groundwater evaporation line is approximately 3.1 (Fig. 8a).

Strontium isotopes
Strontium isotope ratios of surface-and groundwaters in the Cressbrook Creek catchment range from 0.7042 to 0.7119 (Fig. 9), although most samples are within a narrower range of 0.7051 to 0.7078. No measurements of the 87 Sr / 86 Sr ratios of rainwater were conducted for the study area, and as a consequence, the 87 Sr / 86 Sr ratios of rainfall used in this study (Fig. 9a) are based on data from elsewhere in Australia. The 87 Sr / 86 Sr ratios of rainfall are typically similar to modern seawater (0.7092; Dia et al., 1992) near the coast, but they become progressively more radiogenic inland due to the addition of atmospheric dust. Strontium isotope measurements of rainfall from Hamilton, Casterton and Willaura in Victoria (southeastern Australia), which are located approximately 60, 70 and 100 km from the coast respectively, were 0.7094, 0.7097 and 0.7107 (Raiber et al., 2009). In comparison, the rainfall 87 Sr / 86 Sr ratio measured at Woodlawoodlana located approximately 500-600 km inland in South Australia is 0.71314 (Ullman and Collerson, 1994  Creek catchment is approximately 70 km from the eastern coast of Australia (Fig. 1). Assuming a similar increase of the strontium isotope ratios of rainfall with increasing distance from the coast, the 87 Sr / 86 Sr ratios of rainfall in the Cressbrook Creek catchment may be in a similar range to those reported by Raiber et al. (2009), although it is acknowledged that local factors and temporal variability can have a substantial influence. However, the 87 Sr / 86 Sr isotope ratio of rainfall at Cressbrook Creek should not be significantly different to the range presented in Fig. 9a, and any local variations would not affect the hydrological interpretation.

Groundwater residence times
Tritium and 14 C activities have been used to qualitatively assess groundwater residence times in the alluvium and in the surface water of Cressbrook Creek. over 20 years (Morgenstern et al., 2010;Tadros et al., 2014). In Southeast Queensland, the 3 H activities of rainfall have been estimated to be in the range of 1.6-2 TU for the period from 2005 to 2011 (Tadros et al., 2014). The 3 H activity of rainfall is no longer affected by interference from bomb tritium, but is instead controlled by natural cosmogenic production, allowing for a more accurate interpretation of groundwater residence times using a single 3 H measurement (Morgenstern and Daughney, 2012). The 14 C activities of DIC can also provide insight into groundwater residence times and recharge processes. However, the interpretation of 14 C ages is often difficult, because 14 C activities can be altered by geochemical processes that occur in the unsaturated-and saturated zone (Plummer and Glynn, 2013). Nuclear weapons testing further complicated interpretation of 14 C ages in modern samples by increasing atmospheric 14 C activities in the 1950s. The radiocarbon activity of alluvial groundwater in Cressbrook Creek catchment ranges from 81.12 to 104.22 pMC (Table 4). Conventional radiocarbon ages calculated from these data range from modern to BP 1650 years. Two samples (B37 and B83) have modern uncorrected 14 C ages, which correlate well with their relatively high 3 H activities of 1.08 and 1.15 TU, respectively. This confirms that there is a substantial modern groundwater component contained in these groundwaters.
The uncorrected 14 C ages of the samples collected from B57, B36, B18 and B51 are BP 55, 345, 1025 and 1680 years, respectively. However, it should be noted that the 14 C ages have not been corrected for interactions with carbonate minerals. Tritium analyses of the same samples (B57, B36, B18 and B51) indicate that they contain a modern component (i.e. less than approximately 70 years old), with values of 1.02, 0.70, 0.50 and 0.13 TU, respectively.

Hydrochemical facies
Surface-and groundwaters in the upper part of the catchment are generally fresh, with SC values of < 700 µS cm −1 (Table 3; Fig. 3), whereas salinities are moderately higher in the lower catchment. Five hydrochemical facies have been identified based on a visual analysis of major ion proportions (Fig. 5). These hydrochemical facies differ slightly from those presented by King et al. (2014), which were determined by hierarchical cluster analysis (HCA) using major and minor ions and pH. Despite these differences, there are similarities between the hydrochemical facies identified in this paper and the clusters derived using HCA. For example, Hydrochemical Facies 5 of the current paper correlates well with subclusters B2 and B3 of King et al. (2014) and Hydro-chemical Facies 3 shares similarities with subclusters A3 and A1.
Hydrochemical Facies 1 to 3 contain fresh water samples (SC < 1150 µS cm −1 ; Table 3) and samples assigned to these facies have similar concentrations of Ca, Mg and Na (no dominant cation), and low SO 4 concentrations (2.5-62.9 mg L −1 ); therefore, these three groups are mainly distinguished by the relative proportions of Cl to HCO 3 . Hydrochemical Facies 1 is mostly composed of fresh bedrock groundwater samples, but interestingly, it also includes one surface water sample (OCk). This group is characterised by HCO 3 -dominated waters with molar HCO 3 : Cl ratios of ≤ 5. Si concentrations are relatively high (median SiO 2 concentration of 43 mg L −1 ) and low nitrate concentrations (median NO 3 concentration of 0.15 mg L −1 ; Table 6). Hydrochemical Facies 2 and 3 are composed of fresh water samples with slightly higher Cl concentrations than samples assigned to Hydrochemical Facies 1 (49-297 mg L −1 ). Hydrochemical Facies 4 and 5 both contain brackish groundwaters (SC ranges from 1145 to 13 750 µS cm −1 ) with Cl as the dominant anion, but the samples in Hydrochemical Facies 5 have a median NO 3 concentration of 4.0 mg L −1 , compared to those in Facies 4 which have a median NO 3 concentration of just 0.19 mg L −1 (Figs. 5 and 6 and Table 6).

Bedrock groundwater
Bedrock groundwater samples have diverse hydrochemical compositions (Facies 1, 2 and 5; Table 3) and 87 Sr / 86 Sr ratios ( Fig. 9), reflecting the wide range of bedrock types in the study area including granodiorite, basalt, sandstone and shale. Hydrochemical end-members are highly variable due to superimposed processes such as evaporation of water from the unsaturated zone prior to groundwater recharge, transpiration, and mixing from multiple sources. The dominance of HCO 3 for bedrock samples in the upper catchment (Hydrochemical Facies 1) suggests that there are several potential processes that contribute towards the observed patterns of major ion concentrations, including carbonate dissolution, oxidation of organic matter, and silicate weathering. The latter can be assessed using 87 Sr / 86 Sr ratios and silicate stability diagrams.
Groundwaters from the Esk Formation (B229, B103 and B92; Fig. 3) typically have low 87 Sr / 86 Sr ratios (0.7042-0.7062), even though the weathered soils from this formation are comparatively radiogenic (Fig. 9a) with values ranging from 0.7070 to 0.7115 and a mean of 0.7090 (Douglas et al., 2007). This suggests that 87 Sr / 86 Sr ratios of groundwaters from the Esk Formation do not reflect the weathered whole-rock signature, but are instead probably controlled by weathering of plagioclase. Weathering of anorthite (Ca-rich plagioclase) releases 86 Sr (substituted for Ca) into groundwater, but very little 87 Sr is released (McNutt, 2000), resulting in groundwaters with low 87 Sr / 86 Sr ratios. Many other studies have also reported similar observations where groundwa-  Fritz et al., 1992;Richards et al., 1992;Made and Fritz, 1989). This plagioclase dissolution process is supported by geochemical evidence, which shows that Esk Formation soils are rich in smectite (Douglas et al., 2007), and that Ca-rich minerals of the Esk Formation, such as anorthite, are likely to weather to smectite (Fig. 7), whereas minerals that are rich in K, Na and Mg are likely to weather to kaolinite. Therefore, it appears as though silicate weathering is a significant process affecting the major ion concentration of the bedrock groundwaters, particularly in the Esk Formation.

Alluvial groundwaters
Alluvial groundwater evolution is marked by an increase in salinity (Fig. 5), longer groundwater residence times, a decreasing 87 Sr / 86 Sr ratio (Fig. 9b) and higher Cl / HCO 3 ratios (Fig. 5). The more evolved groundwaters in Hydrochemical Facies 4 and 5 have probably been subjected to higher degrees of evapotranspiration. Evaporation processes are evident from stable isotopes measurements, which show that most samples collected during this study are displaced significantly to the right of the Brisbane and Toowoomba MWL (Fig. 8a). This is in agreement with pan evaporation rates that far exceed the average annual rainfall in the catchment (Sect. 2.1). In addition to evaporation, transpiration also appears to be an important control of groundwater salinity in some areas, as documented by elevated Cl and stable isotope signatures that do not show any substantial influence of evaporation (Fig. 8b). However, Mg and Ca concentrations of the samples from Hydrochemical Facies 4 are higher than would be expected from evaporation, based on modelled evaporation curves from fresh water samples from the Upper and lower catchment (Fig. 6).
Similarly, the Na concentrations are lower than expected from the evaporation curve, suggesting that the groundwater composition of samples assigned to Hydrochemical Facies 4 have been influenced by interactions with aquifer materials. As carbonate rocks are absent in the alluvium of this catchment, weathering of silicate minerals appears to be the most likely source of dissolved ions. This is also sup- Note: % CBE = percentage charge balance error. Sub-catchment boundaries are shown in Fig. 3. ported by a moderate correlation between 3 H and the saturation indices (SI) of albite (R 2 = 0.45; Fig. 9d), compared to the weak correlation between 3 H and calcite SI (R 2 = 0.24; Fig. 9e). Furthermore, many of these more evolved waters have Ca / HCO 3 ratios (and Mg / HCO 3 ratios) that are higher than the 1 : 2 molar ratio that could be expected from the dissolution of carbonates alone ( Fig. 6; Appelo and Postma, 2005).
It is likely that this increase in Ca and Mg is augmented by dissolution of mafic minerals such as olivine, pyroxene and anorthite, which are commonly present in basaltic rocks such as those in the mid to upper catchment (Palaeozoic rocks; Fig. 3). Alluvial sediments probably contain detrital material that was eroded off these basalts, providing a source of Ca and Mg for alluvial groundwaters and surface waters in the lower part of the catchment. This is supported by XRD analyses, which show that there are significant amounts of smectite in weathered sediments sampled from Lake Wivenhoe ( Fig. 1; Douglas et al., 2007), and silicate stability diagrams (Fig. 7) demonstrate that the smectite is probably the result of the weathering of Ca-rich minerals such as anorthite.
In contrast to Hydrochemical Facies 4, the samples from Hydrochemical Facies 5 have followed a different evolutionary pathway (Fig. 5): groundwaters that are members of Hydrochemical Facies 5 generally have longer residence times (Table 6), higher Na concentrations (Fig. 5) and a groundwater evolution that more closely follows an evaporative trend (Fig. 6). Nevertheless, the evaporation curve (Fig. 6) indicates that Ca and Mg concentrations are still higher than expected if evaporation alone was the controlling factor, suggesting that the dissolution of silicates is also an important process influencing the chemistry of these waters.

Radiocarbon groundwater residence times
The uncorrected 14 C ages of the samples collected from B18 and B51 are BP 1025 and 1680 years, respectively; however, Table 4. Water isotopic and hydrochemical data for surface-and groundwater samples from the Cressbrook Creek catchment. Saturation indices (SI) for calcite and albite were calculated using PHREEQC (Parkhurst and Appelo, 1999  tritium analyses indicate that this groundwater has a modern component. This discrepancy between the apparent tritium ages and the 14 C ages indicates that the 14 C activity may have been altered by carbonate dissolution, or alternatively, that there has been mixing between an older water component and a younger water component that contains tritium. The Ca : Na ratio of the alluvial groundwaters ranges from 0.19 to 1.00, with an average of 0.54 and the Ca / Na ratio of the samples from B18 and B51 are 0.19 and 0.24. This indicates that significant calcite dissolution is unlikely, as groundwaters that have experienced significant calcite dissolution generally have Ca : Na ratios > 1 (Mast et al., 1990;Leybourne et al., 2006).
Calcite dissolution can also be assessed using the δ 13 C DIC composition, which is affected by interactions with organic materials and the aquifer substrate. The δ 13 C DIC composition of recharging groundwater is largely controlled by the composition of the decomposing plant matter. For plants that use the C 3 photosynthesis, the δ 13 C DIC composition of the soil is usually around −23 ‰, whereas it is likely to be approximately −9 ‰ in areas with C 4 plants (Clark and Fritz, 1997). The study catchment is located in a water-poor area and plant productivity is often limited by the lack of water. Therefore, landholders commonly cultivate plants that use water efficiently, such as those that use the C 4 carbon fixation pathway (e.g. corn and sorghum). However, some droughtresistant plants that use the C 3 carbon fixation pathway (e.g. Lucerne) are also cultivated. Similarly, approximately 74 % of grass species in the Cressbrook Creek region use the C 4 carbon fixation pathway (Hattersley, 1983).
Assuming that approximately 60-90 % of the 13 C is derived from plants that use the C 4 carbon fixation pathway, soil CO 2(g) δ 13 C DIC values would be approximately −15 to −10 ‰. The δ 13 C DIC value will typically increase by around 7.9 ‰ as soil CO 2(g) dissociates to HCO − 3 (at 25 • C; Clark and Fritz, 1997), which will result in groundwater with δ 13 C DIC values between around −7 to −2 ‰. The δ 13 C DIC values at B18 and B51 are −4.4 and −4.9, indicating that there has probably been no significant dissolution of old calcite, and that the uncorrected 14 C ages are valid. This is not unexpected, as the alluvium is composed primarily of components derived from erosion of silicate rocks, and it is unlikely to contain significant amounts of carbonate.

Cressbrook Creek and Cressbrook Dam
Surface water samples from Cressbrook Creek follow an evaporative trend line that intersects the meteoric waterline near the flood-generating rainfall (Fig. 8a). Cressbrook Dam was overflowing into Cressbrook Creek at the time of sampling (Toowoomba Regional Council, 2012), and water from the dam appears to be dominated by depleted heavy rainfall from December 2010 and January 2011. This is not surprising, as the storage volume of Cressbrook Dam was at record low levels (7.5 % of total capacity) in February 2010 (Toowoomba Regional Council, 2014). In addition, rainfall in the Catchment Headwaters and at Cressbrook Dam may be further depleted due to the altitude effect, as the dam is approximately 250 m AHD and the surrounding hills reach elevations of more than 500 m AHD. Stable isotopes were a valuable tool for the identification of episodic recharge in this study. Previous studies have used stable isotopes to link groundwater recharge with high-precipitation rainfall events. For example, Cendón et al. (2010), compared alluvial groundwater isotope signatures with the weighted average isotopic signatures of rainfall events that were greater than 95 mm. However, it is rare to use stable isotopes to assess groundwater recharge from an individual flood event.

Upper catchment
In the upper catchment, recharge to the alluvium is dominated by diffuse infiltration of rainfall rather than channel leakage ( Fig. 10a and b). This is supported by evidence that indicates that the stream is gaining in this part of the catchment, including field observations of groundwater discharge into the stream in the upper catchment, the sustained flow in Cressbrook Creek during years when there was no discharge from Cressbrook Dam and the increase in discharge volume between Cressbrook Dam and CC3 (Fig. 3) at the time of sampling (Sect. 2.1). Groundwater is recharged rapidly in this part of the catchment, based on the low salinity (Fig. 3).
In addition, the comparison of the high groundwater 3 H activities (Table 4) with the rainfall 3 H activities presented by Tadros et al. (2014) for the period 2005-2011 for Southeast Queensland and stream waters analysed during this study confirms that a high component of groundwater consists of very recent recharge. This documents the usefulness of 3 H to assess recharge processes to alluvial aquifers, as previously highlighted by other studies elsewhere (e.g. Cartwright and Morgenstern, 2012;Baudron et al., 2014). Groundwater major ions and stable isotopes from samples collected near the confluence of Cressbrook Creek and Kipper Creek are similar to the surface water sample collected from Kipper Creek (KC1; Fig. 3

), suggesting that Kipper
Creek receives baseflow from the alluvium in the vicinity of KC1. As there was no flow in Kipper Creek in the catchment headwaters at the time of sampling, the creek must have received groundwater baseflow in the Upper Catchment (i.e. near KC1). The stable isotope signature of groundwaters collected from the upper catchment and surface water from Kipper Creek is intermediate to the evaporation trends that originate from the flood-generating rainfall and the longer-term weighted average rainfall value. This suggests that recharge is sourced from the flood and from smaller rainfall events that occurred subsequent to the flood. However, the sample collected from Cressbrook Creek in the upper catchment has a more depleted stable isotope signature than other surface waters or groundwater samples from the upper catchment, probably because water in Cressbrook Creek has a high proportion of isotopically depleted flood runoff and quick flow from Cressbrook Dam (Sect. 5.3.1).
The sample collected from Oaky Creek (OCk; Fig. 3  from the upper catchment, including a sample collected from the granodiorite foothills in the Oaky Creek sub-catchment (B104; Fig. 3). The bedrock appears to have a major impact on the chemical composition of the water in Oaky Creek, probably because the alluvial aquifer is thin and narrow in the Oaky Creek sub-catchment and because the upper layers of granodiorite are highly weathered, and therefore comparatively permeable. This permeable weathered granodiorite probably provides baseflow to Oaky Creek.

Mid to lower catchment
Most groundwaters from the lower part of the catchment also follow the evaporative trend that intersects the meteoric waterline near the flood-generating rainfall of December 2010 and January 2011, indicating that groundwater was recharged rapidly by channel leakage and/or that the flood generated substantially more recharge than other smaller rainfall events. Heavy rainfall events often have depleted stable isotope signatures, as suggested by the depleted signatures of rainfall in December 2010 and January 2011 (Table 5), and observations elsewhere (e.g. Hughes and Craw-ford, 2013). In addition, the most devastating flood to affect southeast Queensland occurred in 1974, and the δ 2 H and δ 18 O values of rainfall during this event were −64.2 and −9.5 ‰, respectively (IAEA/WMO, 2014). In the lower catchment, fresh groundwaters with short residence times, such as those contained in Hydrochemical Facies 2 and 3, are probably recharged rapidly by surface waters (Fig. 10c and d). These sites are generally located close to the creek and it appears as though groundwater quality is significantly improved due to interactions with surface water in this part of the catchment, confirming the observation from King et al. (2014). These groundwater-surface water interactions also appear to affect surface water compositions, as is evident from observed changes in the chemical composition of Cressbrook Creek with distance downstream. This includes (1) an increase in total dissolved salts (Fig. 3); (2) an increase in the apparent water age, as indicated by the 3 H activities at CC1 (Upper Catchment; 1.60 TU) and CC6 (lower catchment; 1.44 TU; Table 4); (3) enrichment of stable isotopes (δ 2 H and δ 18 O; Fig. 8b); and (4) a decrease in the 87 Sr / 86 Sr ratios (Fig. 9a).
Groundwater-surface water interactions are subjected to a large degree of temporal and spatial variations due to the losing/gaining nature of the stream. Cressbrook Creek is generally losing after high-flow events, such as the flood of January 2011, but turns into a gaining stream after this event (Fig. 4). Similarly, the losing/gaining condition of the stream is likely to vary spatially as a result of changes in the streambed elevation (riffles and pools) and groundwater levels (Winter et al. 1998).
The more evolved groundwater samples from the mid and lower catchment (Hydrochemical Facies 4 and 5) were generally collected from alluvial wells that are located further away from the creeks (Fig. 3) and/or where the unsaturated zone is thick (e.g. >10 m). These sites are also located in areas where the alluvium is less permeable, suggesting that infiltrating rainfall from small rainfall events is subjected to a significant degree of evapotranspiration processes during infiltration through the unsaturated zone, and large rainfall events are probably required to generate groundwater recharge. Therefore, it is likely that these more evolved waters are predominately recharged during high-rainfall events, such as those associated with the flooding in January 2011.

Hydraulic connectivity between bedrock and alluvium
Tritium data show that alluvial groundwaters assigned to Hydrochemical Facies 5 have relatively long residence times (B90, B18 and B51, Table 6). In particular, the low 3 H activities from B18 and B51 (0.50 and 0.13 TU, respectively) indicate that older bedrock groundwater could be interacting with the alluvium at these sites. Furthermore, the sample collected from B158, which is screened in both the alluvium and the bedrock, is also included in Hydrochemical Facies 5.
The sample from B90 has a stable isotope signature that indicates a substantial degree of evaporation (Fig. 8), whereas other alluvial samples assigned to Hydrochemical Facies 5 (B51 and B18) are isotopically more depleted. As previously mentioned, alluvial groundwaters assigned to Hydrochemical Facies 5 were probably subjected to significant amounts of evaporation. However, groundwater samples from sites B18 and B51 (Fig. 3) have a relatively depleted stable isotope signature considering their high Cl concentrations (Fig. 8), which suggests that these sites may have received seepage from depleted bedrock groundwater. This was independently confirmed by the use of strontium isotopes, which are ideally suited for the assessment of seepage from basalts and granitic rocks to the alluvium due to their distinguishable signatures resulting from their contrasting mineralogy and the very different ages of the rocks. This study builds on work by Raiber et al. (2009), who used strontium isotopes to investigate interaction of groundwater with basalts and granitic rocks in southwestern Victoria, Australia. The groundwater sample from B18 has a radiogenic 87 Sr / 86 Sr signature similar to groundwater sampled from the granodiorite, which forms the bedrock at this site, and sample B51 has a low 87 Sr / 86 Sr ratio similar to the Esk Formation samples (Fig. 9a). Also, the 14 C groundwater ages of samples from B18 and B51 are greater than BP 1000 years, but there is detectable tritium in these samples, indicating that the water is less than approximately 100 years old. This discrepancy is consistent with mixing of old bedrock groundwater with younger alluvial groundwater.
Overall, the isotopic evidence (groundwater 14 C, δ 13 C, stable isotopes and 87 Sr / 86 Sr ratios) confirms that the aquifer at sites B18 and B51 receives seepage from the underlying bedrock. Interestingly, the only other two samples with 87 Sr / 86 Sr ratios below the 99 % confidence interval (Fig. 9a) are the samples from B36 and B57, which were also collected from monitoring bores overlying the Esk Formation. Furthermore, apart from B18 and B51, they are the only other two samples with non-modern uncorrected 14 C ages, and they have relatively depleted stable isotope signatures, suggesting that the alluvial aquifer at these sites has probably also received seepage from the underlying bedrock aquifer. Moreover, there is a strong correlation (R 2 = 0.94) between 3 H and 14 C activities (Fig. 9c), which suggests that the samples from B57 and B36 have been affected by similar hydrological processes (i.e. bedrock seepage) as the samples from B51 and B18.

Conclusions
This study outlines the benefits of the simultaneous application of multiple environmental isotopes ( 2 H, 18 O, 87 Sr / 86 Sr, 3 H and 14 C) in rainfall, groundwater and surface water in combination with a comprehensive hydrochemical assessment. The aim was to study the influence of a flood on groundwater recharge and to assess the hydrological connectivity of an alluvial aquifer system with associated streams and underlying highly diverse bedrock aquifers.
Groundwater evolution is largely controlled by silicate dissolution and evapotranspiration processes, as demonstrated by the silicate stability diagrams, theoretical evaporation curves and saturation indices. In the upper catchment, rainfall is quickly recharged through relatively coarse-grained alluvial sediments. Conversely, rainwater infiltrates more slowly in the mid and lower catchment, particularly in the floodplain distal to Cressbrook Creek, as indicated by the lower tritium and 14 C values and the elevated salinity. In contrast, surface water leakage to the alluvial aquifer is an important mechanism for maintaining groundwater quality and for the generation of recharge in the lower part of the catchment.
The flood-generating rainfall in 2011 was isotopically more depleted (δ 2 H and δ 18 O) than the long-term weighted average, and groundwater from the lower part of the catchment plots along an evaporative trend line that intersects the meteoric waterline near this depleted, flood-generating rainfall of December 2010 and January 2011. This confirmed that the flood events of January 2011 generated significant recharge, whereas infiltrating water from smaller rainfall events is subject to evapotranspiration, especially in the lower part of the catchment where the unsaturated zone is relatively thick and the permeability is low. Recharge from episodic flooding is probably important in other similar settings where low-permeability sediments are incised by stream channels. Groundwater in the upper catchment follows an evaporative trend initiated from rainfall that is intermediate to the long-term weighted average rainfall and the "flood rainfall". The floods of 2011 also generated significant recharge in this part of the catchment. However, as the evaporative trend is initiated from a more enriched rainfall signature (i.e. closer to the long-term weighted average), it appears likely that smaller rainfall events also generate groundwater recharge here, probably due to the more permeable and thinner soil material in this part of the catchment. The study clearly demonstrated the value of time series rainfall stable isotope data for the identification of hydrological processes such as aquifer recharge and the generation of baseflow resulting from flooding.
The 87 Sr / 86 Sr ratios were used to identify bedrock seepage to the alluvium at several locations. This conclusion was supported by the 3 H and 14 C data, which show that the alluvium contains a mixture of older, bedrock-derived groundwater and more recently recharged groundwater. The connectivity between the alluvium and the bedrock is likely to be spatially and temporally variable.
The complementary use of multiple isotopes and hydrochemistry of rainfall, groundwater and surface water enabled an effective assessment of hydrological processes throughout the catchment, including recharge of the alluvial deposits from surface water flows and variable bedrock aquifers, recharge specifically from flood events and an understand-