The influence of oxygen isotope exchange between CO2 and H2O in natural CO2-rich spring waters: implications for geothermometry

Oxygen isotope ratio ( d 18 O) value deviations from the Meteoric Water Line with no signi ﬁ cant change in the hydrogen isotope ( d 2 H) composition have been reported in naturally occurring CO 2 -rich waters from around the world. Here we review the effects of oxygen isotope exchange with CO 2 , high temperature equilibration with bedrock minerals and mineral dissolution and precipitation reactions on the CO 2 -rich water isotopic composition. We present two case studies from Daylesford (Australia) and Pah Tempe (Utah, USA) mineral springs, where we use a numerical geochemical modelling approach to resolve the in ﬂ uence of low temperature water-rock interactions and CO 2 equilibration to the oxygen isotope ranges observed in the mineral waters. In both cases, we ﬁ nd that mineral dissolution e precipitation reactions are unlikely to have a signi ﬁ cant effect on the groundwater isotopic compositions, and that the observed d 18 O values in natural CO 2 springs can be simply explained by equilibrium fractionation between water and free phase CO 2 . Traditionally, the interaction of CO 2 and water in a natural CO 2 -rich groundwater setting has only been associated with water 18 O depletion and this is the ﬁ rst study to consider 18 O enrichment. We establish that in a natural setting, CO 2 and water equilibration can result in water 18 O depletion or enrichment, and that the change in the oxygen isotope composition ultimately depends on the initial CO 2 and water d 18 O values. Our new conceptual model therefore provides a mechanism to explain water 18 O enrichment at ambient temperatures. This ﬁ nding is critical for the use of d 18 O in groundwater geothermometry and for the interpretation of natural water circulation depths: we argue that in some cases, natural waters previously interpreted as geothermal based on their oxygen isotope composition may actually have acquired their isotopic signature through interaction with CO 2 at ambient temperatures.

The influence of oxygen isotope exchange between CO 2 and H 2 O in natural CO 2 -rich spring waters: Implications for geothermometry

Introduction
The stable isotope ratios in groundwater are useful indicators of a wide range of geological conditions associated with groundwater reservoirs and mineral springs. Applications include interpretation of water origin (Harris et al., 1997;Ziegler, 2006;Demlie and Titus, 2015), residence times (Vuataz and Goff, 1986;Hearn et al., 1989), migration pathways and mixing trends (Hearn et al., 1989;Siegel et al., 2004;Wilkinson et al., 2009;Delalande et al., 2011), temperature and circulation depths (Ziegler, 2006;Nelson et al., 2009), fault and fracture permeability to fluids (Cer on et al., 1998;Losh et al., 1999;Lions et al., 2014), local rainfall variations (Burns and Matter, 1995) and paleoclimates (Hays and Grossman, 1991). These geological interpretations rely on the identification of fundamental natural processes controlling kinetic and equilibrium fractionation of water stable oxygen and hydrogen isotopes (Fig. 1). Here we investigate the effect of water interaction with CO 2 and its impact on the isotopic composition of natural waters. In particular, we investigate changes in the oxygen isotope ratio (d 18 O) values of water independent of changes in hydrogen isotope ratios (d 2 H).
The main processes that result in water 18 O depletion or enrichment without affecting hydrogen isotope ratios are oxygen isotope exchange with CO 2 and isotope exchange with bedrock minerals, either through low temperature mineral dissolution and precipitation, or diffusion at high temperatures (Fig. 1). Water 18 O enrichment relative to the Meteoric Water Line (MWL) with no change in the hydrogen isotope ratio has been traditionally associated with oxygen isotope exchange with bedrock minerals or water-steam separation in geothermal conditions (e.g. Clayton and Steiner, 1975;Matsuhisa et al., 1979;Giggenbach, 1992). Due to the lack of other reported water 18 O enriching mechanisms it has become common practice to interpret 18 O enrichment in water as evidence for geothermal conditions (e.g. Cer on et al., 1998;Nelson et al., 2009). In contrast, oxygen isotope exchange with CO 2 has been associated with water 18 O depletion (Clark and Fritz, 1997;D'Amore and Panichi, 1987). However, recent applications of oxygen isotopes to monitor injected CO 2 in Carbon Capture and Storage (CCS) reservoirs have demonstrated that the water isotopic composition after CO 2 injection is primarily dependent on the oxygen isotope ratios of pre-injection water and CO 2 and of the degree of CO 2 saturation in the water and gas phases in the reservoir pore space (Kharaka et al., 2006;Serno et al., 2016). Applying this new knowledge to naturally occurring CO 2 -rich mineral springs, we re-evaluate previous interpretations of d 18 O and d 2 H geochemistry and propose a new conceptual model to explain the global occurrence of 18 O shifts in CO 2 -rich waters.
In two case studies from Daylesford (Australia) and Pah Tempe (Utah, USA), we investigate the effects of mineral reactions and oxygen isotope exchange with CO 2 to the observed water oxygen isotope composition. We propose a method to assess the relative contributions of these two mechanisms, which can be applied to waters with elevated CO 2 concentrations in both natural and engineered settings.

Global and local meteoric water lines
The ratios of stable oxygen (d 18  values in global precipitation reflected by the Global Meteoric Water Line (GMWL), first defined by Craig (1961) and refined by Rozanski et al. (1993): The slope of the line is produced by equilibrium Rayleigh fractionation as precipitation is successively removed from the vapour phase when it condenses, leaving the residual water vapour progressively depleted in 18 O and 2 H. The intercept of the line is controlled by kinetic fractionation during evaporation of seawater. Variations in humidity and temperature affect the slope and intercept of the line, and produce different Local Meteoric Water Lines (LMWL) for specific locations (Dansgaard, 1964;Clark and Fritz, 1997).

Meteoric water stable isotope change due to natural processes
Various natural processes may deviate the stable isotope ratios in reservoir waters from equilibrium values on the MWLs (Fig. 1). During evaporation, lighter isotopes enter the vapour phase, whereas in condensation heavier isotopes are preferentially incorporated into the condensate. Consequently, water vapour is depleted in 18 O and 2 H whereas the remaining water is enriched. Similarly, temperature-dependent kinetic fractionation occurs during steam loss above the boiling temperature, which produces 18 O enrichment. (Clark and Fritz, 1997). Fractionation between degassing H 2 , H 2 S, CH 4 , and water in active magmatic systems leads to enriched water d 2 H values without an effect on d 18 O (Richet et al., 1977). CO 2 -rich waters are often characterised by a horizontal deviation from the MWL (e.g. D'Amore and Panichi, 1987; Pauwels et al., 1997; Cartwright et al., 2000). CO 2 presence in the system can lead to water d 18 O changes by either: i) Promoting low temperature primary mineral dissolution and secondary mineral precipitation reactions, preferentially consuming 18 O. ii) Equilibrium oxygen isotope exchange between CO 2 and water.
Both of these processes affect only the water d 18 O values while d 2 H values remain unchanged, unless water-rock reactions involve extensive precipitation of H-rich clays (D'Amore and Panichi, 1987). Additionally, diffusive equilibrium oxygen isotope exchange with bedrock minerals at high temperatures produces 18 O-enriched waters. Here, we review and assess the relative contributions of these three processes (Fig. 1) to the observed water d 18 O values of a global dataset of natural mineral waters.

Low temperature dissolution e precipitation reactions
The temperature-dependent oxygen isotope equilibrium fractionation factor between water and a precipitated mineral leads to a preferentially 18 O-enriched mineral phase and lower d 18 O values of the water. The extent of this enrichment depends on the strength of mineral crystal lattice bonds (Zheng, 2011). Under equilibrium conditions, secondary minerals such as clays and carbonates preferentially incorporate more 18 O during precipitation and hence become enriched relative to the water (Compton et al., 1999;Kloppmann et al., 2002). Clays also incorporate water molecules in the isotopically depleted intra layer, which leads to a positive d 2 H value shift in the remaining pore water (Sheppard and Gilg, 1996).
This process can significantly alter the water d 18 O if the fraction of oxygen involved in the reactions is sufficiently high (D' Amore and Panichi, 1987). Mineral dissolution and precipitation are particularly important in CO 2 -rich waters, which are often associated with primary silicate hydrolysis and enhanced clay production rates (e.g., Watson et al., 2004;Kampman et al., 2014).

Diffusion related equilibrium oxygen isotope exchange with minerals
Meteoric water circulating at depth is depleted in 18 O compared to rock-forming minerals. Oxygen isotope exchange between the fluid and solid phases via diffusion gradually moves the two phases towards equilibrium at a rate controlled by the water temperature. This is important in geothermal systems where the disequilibrium is higher because the fractionation factor between water and minerals is low. In many cases, deep groundwaters are enriched in 18 O respective to the GMWL, with little change in d 2 H values as a consequence of equilibrium oxygen isotope exchange with bedrock minerals that are commonly low in hydrogen (Clark and Fritz, 1997). However, common rock-forming minerals such as feldspar, mica and quartz require heating to temperatures above 250 C to achieve this oxygen isotope exchange (D'Amore and Panichi, 1987). Hence the effect is only observed in geothermal conditions (Friedman and O'Niel, 1977).

Equilibrium oxygen isotope exchange with CO 2
Equilibrium oxygen isotope fractionation occurs during redistribution of isotopes between two or more compounds with forward and backward reactions proceeding at equal rates. At isotopic equilibrium heavier isotopes preferentially concentrate in the phase with stronger bond constants (Young et al., 2002). In the case of CO 2 and water, CO 2 has stronger bonds than water and thus equilibrium exchange results in the CO 2 phase being enriched in 18 O. The fractionation factor (a) associated with equilibrium exchange reactions between CO 2 and water is expressed as: The fractionation factor is inversely correlated with temperature and is therefore particularly important for low temperature waters (Bottinga, 1968). In most natural systems the oxygen isotope equilibrium between CO 2 and water is predominantly influenced by the initial water d 18 O value as in the majority of low pressure systems water represents the greater source of oxygen. However, in cases where CO 2 represents a major source of oxygen, the isotopic composition of water may be influenced by CO 2 (e.g., Kharaka et al., 2006;Serno et al., 2016).

Overview of global CO 2 -rich water isotopic compositions
Observed changes in the oxygen isotope geochemistry of global CO 2 -rich waters have been associated with isotopic equilibrium exchange between natural free phase CO 2 and formation waters (Fig. 2). Examples from the literature include both thermal and cold springs with temperatures similar to those of ambient groundwater. Water 18 O depletion and enrichment without a change in d 2 H has been observed in cold springs in the Valles Caldera-Southern Jemez Mountains in New Mexico, USA (Vuataz and Goff, 1986), along the Bongwana gas fault in South Africa (Harris et al., 1997) and in shallow boreholes in the Mont-Dore and Montmiral regions of the Massif Central, France (Casanova et al., 1999;Humez et al., 2014;Pauwels et al., 1997Pauwels et al., , 2007. The Alto Guadalentín groundwater aquifer in southeast Spain showed a similar shift in d 18 O which has been associated with the manifestation of CO 2 from greater depth due to overexploitation of the aquifer, and oxygen isotope equilibrium exchange between the CO 2 gas and reservoir water (Cer on et al., 1998;Cer on and Pulido-Bosch, 1999). Geochemical differences in various northern Portuguese CO 2 -rich mineral waters are reportedly caused by water-CO 2 isotopic equilibrium exchange (Marques et al., 2000). A majority of the bubbling pools in the central Italian volcanic region, and some hydrothermal waters from Sicily, exhibit a d 18 O value deviation from the Local Meteoric Water Line (LMWL), which is likely the result of oxygen isotope exchange between the meteoric water and CO 2 (Cinti et al., 2011).
A comparative study of thermal and cold spring waters in the Poroto-Rungwe region in Tanzania revealed that springs with observed sustained CO 2 flux show water 18 O depletion of up to À8‰ relative to the LMWL, while springs with relatively lower and episodic bubbling gas emanations are isotopically similar to surface meteoric water. Water 18 O depletion was negatively correlated with temperature (Delalande et al., 2011), consistent with experimental observations indicating fractionation increase at low temperatures (Bottinga, 1968). Water 18 O depletion caused by mantle-derived CO 2 interaction with cold springs is also observed at the Snaefellsnes Peninsula in Iceland. A number of hot springs from the same region are 18 O enriched, although this is interpreted as a result of isotopic exchange with bedrock minerals at geothermal temperatures (Thomas et al., 2016). Finally, evidence from Hofsstadir (Iceland) shows that the effect of oxygen isotope exchange can be preserved in water after the gas has leaked out of the system. Stagnant geothermal water pools with low CO 2 contents are reportedly depleted in 18 O due to an episode of CO 2 flux in the past (Kristmannsd ottir and Sveinbjoernsd ottir, 2012). Hence, this global compilation of observed CO 2 -rich water oxygen isotope composition clearly shows that 18 (Fig. 2c). The mechanism of water 18 O enrichment at Pah Tempe is currently uncertain. Nelson et al. (2009) suggested equilibrium isotope exchange with bedrock minerals at temperatures above 150 C on the basis of the lack of other plausible mechanism for water 18 O enrichment. However, evidence for >3e5 km deep faults providing circulation pathways is limited (Nelson et al., 2009) and the temperature of the water discharging at the surface ranges from 39 to 41 C (Dutson, 2005). Further, circulation temperatures calculated from other conventional geothermometry techniques indicate lower temperatures: 70e75 C using conductive and adiabatic silica and 37e39 C using chalcedony silica geothermometers (Dutson, 2005) and up to 80 C using quartz and Na-K-Ca-Mg geothermometer (Budding and Sommer, 1986).

Daylesford springs, Australia
Daylesford mineral springs are located in the Central Highlands of Victoria, south east Australia (Fig. 3a). There are more than 100 low temperature CO 2 -rich springs in the area, which have been exploited historically for drinking and recreational purposes. Springs flow in a fracture-dominated aquifer through an Ordovician turbidite sequence altered to greenschist facies and discharge into topographic lows such as streambeds. The depth of circulation is unknown but historical records report spring water in mines at a maximum depth of 1.6 km (Shugg, 2009). The aquifer is overlain by Newer Volcanic basalts, active from 4.5 Ma to 5000 a (Boyce, 2013). The spring waters contain excess dissolved carbon (Weaver et al., 2006) and actively degasses at surface discharge points. CO 2 is reportedly mantle-sourced, based on their close proximity to the eruptive centres (Lawrence, 1969), 3 He/ 4 He gas data (Chivas et al., 1983), and gas carbon isotope values (d 13 C; Cartwright et al., 2002). Spring waters are Na-HCO 3 type with a significant solute composition variation between individual springs (Weaver et al., 2006).

Pah Tempe springs, Utah, USA
The Pah Tempe springs discharge in the damage zone of the currently active Hurricane fault at Timpoweap Canyon, Utah, USA (Fig. 3b). The Hurricane fault has a total displacement of up to 3000 m and a 200 m wide core (Biek, 2003). All springs discharge at the eastern damage zone into the Virgin River Canyon. The stratigraphic sequence includes Cretaceous sandstones, siltstones and shales, and carbonaceous Jurassic e Triassic sediments and evaporites (Biek, 2003;Nelson et al., 2009). Free-phase CO 2 is actively degassing at the surface, forming prominent bubble trains, along with mineralised water at an elevated temperature of 40 C. The spring waters are high in Na þ , Cl À , Ca 2þ and SO 2À as well as dissolved CO 2 (Dutson, 2005). Helium isotope and d 13 C analysis indicate carbonate thermal alteration as a primary CO 2 source (Nelson et al., 2009).

CO 2 sampling
CO 2 samples from Daylesford and Pah Tempe springs were collected using a funnel placed over a bubbling vent and connected by plastic hosing to a copper tube. After purging air from the sampling line by allowing gas to flow for 5e10 min, the copper tubes were sealed by two metal clamps at either side creating a helium leak tight cold weld. At Daylesford water samples were collected via hand pumping the mineral water wells and from stream waters at gas discharge locations where CO 2 samples were collected. Waters were filtered through 0.45 mm pore-size filters and filled into Nalgene bottles (with no headspace) or vacutainers. Samples were stored in a cooler until analysis to avoid evaporation. The temperature of the water which the CO 2 was bubbling through was measured in the field using a Hannah Instruments HI991300 Portable Waterproof pH/EC/TDS Meter. Temperature accuracy is ±0.5 C.

Stable isotope analysis
The d 18 O values of exsolved CO 2 gas samples from Pah Tempe and Daylesford were measured at the Scottish Universities Environmental Research Centre (SUERC). Gas samples were released into an ultra-high vacuum extraction line. Two aliquots of the gas were trapped into glass ampules for stable isotope analysis, using liquid nitrogen and sealing the ampule with a blowtorch after pumping away the other volatile gases. Samples were analysed using a VG Isotech Sira 2b mass spectrometer with typical un-

Numerical simulation of water-rock reactions
Numerical simulations allow the quantification of the mass transfer between the solid and fluid phases during dissolution and precipitation reactions at elevated CO 2 pressures. The amount of oxygen liberated from dissolving minerals and precipitated in secondary minerals can be compared to the total oxygen in the solution to assess the contribution of water-rock reactions to the overall water oxygen isotope signature.
Mineral dissolution and precipitation reactions are simulated using the geochemical modelling software PHREEQC (Parkhurst and Appelo, 1999) with the WATEQ4F database (Ball and Nordstrom, 1991). The numerical simulations solve a set of nonlinear mass balance equations using thermodynamic constants defined in the database.
Simulations require a set of reactive primary and precipitating secondary minerals, which are chosen according to the saturation indices of the dissolved species in groundwater calculated in PHREEQC and the geology of the local areas. The main goal of the simulations was to reproduce the observed geochemistry of the spring waters, published in Weaver et al. (2006) for Daylesford springs and Dutson (2005) for Pah Tempe springs, and to quantify the oxygen isotope transfer between the solid and fluid phases. There are multiple reaction pathways to achieve the observed water chemistry but the model seeks to represent the simplest solution using the most likely reactive phases based on the geological setting.
4.3.1. Model assumptions 4.3.1.1. Initial groundwater and CO 2 dissolution. Equilibrium dissolution and precipitation reactions are modelled in a closed system considering reactive species in 1 L of water. Average modern local groundwater is used as a starting solution (Table 1). CO 2 partial pressure (pCO 2 ) is calculated from measured pH and alkalinity (as HCO 3 À ) values, so mineral reactions are modelled at ambient pressures and average recorded spring discharge temperatures (25 and 40 C for Daylesford and Pah Tempe, respectively). This method provides the best fit to the measured DIC contents. Alternatively, reactions could be modelled at depth with higher pCO 2 and subsequent dilution at the surface. This would allow dissolution of more stable minerals in the beginning (May 2005) but was found not to have a significant effect on the overall mass balance.
4.3.1.2. Mineral reactions in Daylesford, Australia. Major cation concentrations and bicarbonate contents display a positive correlation, indicating that mineral dissolution is proportional to acid neutralisation This is a common feature in Na-Ca-HCO 3 type waters produced by acid groundwater dissolution of silicates (May 2005). These trends infer that the bulk reaction is limited by the kinetics of primary mineral dissolution and that the system is not in equilibrium. SiO 2 does not show correlation with HCO 3 À , suggesting secondary silica precipitation.
The model assumes reactions with common minerals in the Ordovician Castlemaine turbidite sequence and Newer Volcanics intrusions. Albite and chlorite were chosen as source rocks for Na, Mg and Fe in the water based on XRD analysis of average Ordovician Castlemaine turbidites collected in the Ballarat area (Bierlein et al., 1999). The study compares unaltered Ordovician rock with that altered by hydrothermal fluids. Two other major phases, muscovite and quartz, are stable at hydrothermal conditions, and therefore are assumed to be unreactive in CO 2 -water system in the model. Ca and small amounts of K are sourced from feldspars in mafic Ca-rich and trachytic lavas from the Newer Volcanic sequence (Price et al., 2003), and are modelled as anorthite and adularia. Dissolution of minor amounts of sulfates (melanterite, barite) contributes SO 4 , Fe and Ba to the system, consistent with redox values measured by Weaver et al. (2006). Other trace elements (Sr, Mn) are sourced from carbonate dissolution (rhodochrosite, siderite). Secondary minerals are allowed to precipitate to equilibrium were kaolinite, amorphous silica and Mg-carbonates.
4.3.1.3. Mineral reactions in Pah Tempe, Utah, USA. The main model assumptions are based on spring water geochemistry interpretation by Dutson (2005). There is little variation in solute contents between samples collected at different discharge sites (Dutson, 2005), suggesting springs emerge from a single aquifer. The average molar ratios for Na/Cl and Ca/SO 4 are 1.08 and 0.98, indicating halite and gypsum dissolution, which could be sourced from the Triassic sequence (Biek, 2003;Dutson, 2005). As gypsum accounts for most of the Ca contents, carbonate dissolution is unlikely. Minor amounts of Mg and K are introduced by silicate dissolution, modelled as phlogopite. Precipitation of amorphous silica provides a sink for Si.  (Table 2).

Low temperature mineral precipitation e dissolution reactions
Numerical simulations of equilibrium dissolution and precipitation of primary and secondary phases at fixed CO 2 partial pressures (0.8 atm in Daylesford and 0.79 atm in Pah Tempe) produce geochemical compositions that closely match the reported measurements of Na-HCO 3 and Na-Cl-HCO 3 waters from Daylesford and Pah Tempe (Fig. 4). In Daylesford, the modelled Na, Ca, K, SO 4 and trace element (Fe, Sr, Mn and Ba) contents are fixed by the defined amount of primary dissolving minerals, while Mg and Si concentrations are controlled by equilibrium secondary mineral precipitation. For Pah Tempe springs, Na, Cl, Mg and K contents are controlled by fixed amounts of mineral dissolution, Ca and SO 4 are limited by the gypsum solubility, and Si is controlled by the kaolinite and secondary silica precipitation. Both models overestimate Si contents. The exact precipitating silica polymorphs are unknown and the thermodynamic constant for amorphous silica used in the simulation may not be precise. The relative amounts of dissolving minerals are not proportional to the bulk rock composition (as reported by Bierlein et al., 1999) suggesting that carbonic acid alteration is limited by mineral dissolution kinetics. The modelling approach for which fixed amounts of minerals are dissolved to match the observed element concentrations effectively eliminates the uncertainties associated with predicting individual mineral dissolution rates.
The total amount of phases that react are summarised in Table 3. In both cases, mineral reactions liberate more oxygen to the system than remove via precipitation. However, the total amount of oxygen involved in both types of reactions represents only 0.11% and 0.087% of total oxygen atoms in the water in Daylesford and Pah Tempe, respectively.

Water e rock reaction influence to water oxygen isotope composition
The effect of mineral dissolution and precipitation on the water oxygen isotope composition depends on the d 18 O values of the water, dissolving and precipitating phases, and the relative ratios between the solid and fluid phases. This relationship can be expressed by a simple mass balance model: Where X o is the relative fraction of oxygen in the phase, and d 18 O i and d 18 O f are the initial and final oxygen isotope ratios in mineral (m) and water (w). The mass balances obtained from numerical simulations (Table 3) indicate that mineral reactions account for dissolution of 0.7 g and 0.8 g, and precipitation of 0.4 g and 0.04 g of oxygen per 1 L of water in Daylesford and Pah Tempe, respectively. This represents~0.1% of the total oxygen in the water.  (5e40‰) does not exceed 0.01‰, which is below the analytical sensitivity and therefore has a negligible influence on the d 18 O.
It is also important to consider mineral reactions that achieve equilibrium quickly and may not be evident in the water geochemistry, such as primary carbonate dissolution and secondary carbonate precipitation. A theoretical scenario of carbonate dissolution with low d 18 O values formed at high temperatures and re-precipitation at low temperatures would result in a water 18 O depletion of 0.5‰, requiring progressive reworking of 200 g of carbonates per 1 L of water. A recent study by Sterpenich et al. (2009) demonstrated that less than 1% by mass of an oolitic limestone dissolved due to interaction with CO 2 -saturated water under extreme experimental conditions (150 bar, 80 C). These results The mineral water geochemistry in Daylesford and Pah Tempe is controlled by the primary mineral dissolution and secondary mineral precipitation accelerated by elevated CO 2 partial pressures. However, the fraction of oxygen in these reactions is too small to influence the oxygen isotope ratio of the water body. Consequently, water-rock reactions cannot account for water d 18 O deviations from the LMWL in both case studies due to their low salinity. However, the method could be applicable to more saline formations where mineral reaction may liberate enough oxygen to alter water d 18 O values, such as deep basement fluids and hypersaline brines. Depletion of 18 O in saline brines (up to 250 g/L) in Fennoscandian and Canadian Shields have been previously interpreted as a result of low temperature water-rock reactions (Blomqvist, 1990;Frape and Fritz, 1982).

CO 2 -water oxygen isotope exchange influence on water isotopic composition
Our geochemical modelling results clearly show that low temperature mineral reactions can be excluded as a significant source of oxygen to the waters at Daylesford and Pah Tempe. Hence these two case studies, along with the global compilation of stable isotope values from CO 2 -rich springs, provide robust evidence that CO 2 -water oxygen isotope equilibrium exchange in the subsurface can result in 18 O depletion and enrichment in CO 2 -rich spring waters compared to the MWL.
The amount of CO 2 required to achieve the d 18 O H2O change observed in Daylesford (À1.43‰) and Pah Tempe (1.68‰) can be estimated using the conceptual model developed by . The magnitude of the shift in d 18 O H2O relates to the fraction of CO 2 in the system. The extent to which CO 2 can change the oxygen isotope composition of reservoir water depends on the: This relationship is expressed in equation (6) where 18 O f H2O is the final oxygen isotope composition of water : values associated with mantle degassing. The first one (8‰) represents a 'minimum value' scenario of CO 2 degassed from a volcanic source whereby large amounts of mantle-sourced gas ascends through the crust without significant interaction with fluids other than the mineral springs. However, d 18 O is likely to change due the interaction with subsurface fluids with increasing migration or residence time in a natural trap. The second scenario represents mantle CO 2 after interaction with subsurface fluids and uses d 18 O value of 19.1‰ measured in Caroline CO 2 field in Mt.Gambier, South Australia (Chivas et al., 1987), which is associated with the same period of volcanic activity as the eruptive centres in Daylesford. A d 18 O i CO2 value of 30‰ represents an average value for thermogenic CO 2 from a global range (Bindeman, 2008). Equilibration with these potential values for mantle-derived CO 2 at Daylesford would require the fraction of oxygen sourced from CO 2 in the system ðX o CO2 Þ to be 5% and 10% for average mantle and Caroline field CO 2 , respectively, to explain the maximum observed d 18 O deviations. The maximum d 18 O shift observed in Pah Tempe can be explained by equilibration with the average thermogenic CO 2 when X o CO2 is 30%. This amounts to 60e120 g and 366 g of CO 2 per every litre of water in Daylesford and Pah Tempe, respectively.
This simple model uses a closed system two-component mixing approach. In reality, both CO 2 and water move through the system at different rates. If CO 2 moves through a relatively stagnant water body at a continuous rate and degasses at the surface, the calculated X o CO2 ranges represent the amount of CO 2 that the water has interacted with, rather than the amount of CO 2 currently present in the system. Therefore, these values can be taken as a maximum estimate.
The variation of oxygen isotope ratios in MORB-type igneous rocks are between 5.5 and 5.9‰, rhyolitic magmas have values between 5.8 and 6.5‰ and the overall range of various measured igneous lithologies is 4e12‰ (Fig. 5 inset) (Bindeman, 2008). CO 2mineral isotopic enrichment factors measured at melting temperatures range between 2 and 6‰ (Matthews et al., 1994), giving the overall range of volcanic degassing related d 18 O i CO2 values of 6e18‰ with MORB-like signatures at the lower end of the spectrum. In contrast, d 18 O values of sedimentary and carbonate rocks range between 8 and 32‰. CO 2 -mineral isotopic enrichment factors span from 2 to 11‰ depending on temperature and are at the higher end of this spectrum in temperatures relevant to thermogenic gas generation (Zheng, 1999;Zhao and Zheng, 2003 values, as demonstrated using the case studies from Daylesford and Pah Tempe.

Differences between observed and theoretical enrichment factors
We combine newly obtained gas d 18 O CO2 and water d 18 O H2O measurements with those previously published in Dutson (2005) to calculate the expected oxygen isotope enrichment factor ðε CO2ÀH2O ) (Bottinga, 1968) for individual springs (Table 4). For Pah Tempe springs calculations, we use an average temperature for springs where data is not available and d 18 O CO2 value of 26.3 ± 0.9‰, which is an average of two measurements reported in this paper and one measurement of 27.1‰ reported in Dutson (2005). Temperature-dependent oxygen isotope enrichment factors ε CO2ÀH2O calculated for the individual springs are larger than Table 3 Summary of dissolving and precipitating minerals in Daylesford and Pah Tempe springs, and associated oxygen concentrations contributing to the water. Oxygen sourced from mineral reactions is compared to the total oxygen atoms in 1 L of water (55.6 mol/L). the observed difference between d 18 O of water and CO 2 (D) in all cases except for Taradale spring in Daylesford. The difference between the observed and theoretical enrichment factor (D -ε) is 0.2‰ and À1.9‰ for two Daylesford springs and 2 ± 0.4‰ for Pah Tempe. Potential reasons for this apparent dis-equilibrium include the effects of salinity, partial equilibrium during CO 2 ascent and kinetic fractionation on bubble surfaces. These mechanisms have opposing isotopic effects: bubble formation on the surface leads to lower apparent CO 2 -water fractionation factor, while salinity and partial equilibration can either decrease or increase it.

Water salinity
High salinity waters display slightly altered ε CO2ÀH2O depending on temperature and total dissolved solids (TDS). Truesdell (1974) and Becker et al. (2015) reported ε CO2ÀH2O decreasing by 1‰ in laboratory experiments using NaCl solutions of up to 250 g/L. L ecuyer et al. (2009) reported an increase in fractionation factor by up to 0.5‰ for 250 g/L KCl and sea salt solutions, which highlights the fact that different types of ions may have opposing effects. Given the relatively low salinities at Pah Tempe (8 g/L) (Dutson, 2005) and Daylesford (5 g/L) (Cartwright et al., 2002), this effect should be negligible.
6.3.2. Partial equilibration during CO 2 ascent to the surface CO 2 -water isotopic equilibrium established locally or at a certain depth may be affected by the kinetics of two-phase fluid flow as CO 2 ascends to the surface. The extent of mixing achieved by CO 2 and water depends on the nature of interaction at depth. The observed water d 18 O shift requires large quantities of free phase CO 2 interacting with the water. There is significant heterogeneity associated with the two-phase CO 2 and water flow through the subsurface due to pressure and temperature effects on CO 2 physical properties. CO 2 and water may interact as transient, dispersed or separated two-phase flows at varying rates, depending on bedrock properties and interfacial tensions between the two phases (Plampin et al., 2014;Roberts et al., 2015). As pressure and temperature decrease at shallow levels of the subsurface, increased CO 2 buoyancy provides a driving force to migrate at a faster rate. CO 2 may achieve local equilibrium with surrounding water at depth where the flow rate is relatively low and where the isotopic signature is preserved after rapid ascent to the surface. Consequently, the remaining water will re-equilibrate and display a smaller d 18 O shift than expected if in equilibrium with the sampled CO 2 . This may be the case in Pah Tempe springs where the CO 2 flux is strong and sustained -measured d 18 O CO2 may represent equilibrium with water enriched in 18 O by up to 2‰ more than measured at the surface. Degassing in Daylesford is much more diffuse and episodic, allowing more time for equilibration with surface water, so partial equilibration has a less significant effect.

Kinetic fractionation on bubbles
An alternative or additional mechanism, which may deviate d 18 O CO2 from equilibrium, is kinetic fractionation during diffusion of dissolved species towards gas bubbles at the surface. Mass control on diffusivity leads to preferential uptake of 12 C and 16 O during bubble formation (Affek and Zaarur, 2014). If time is not sufficient for re-equilibration, the fractionation factor between CO 2 and water will be lower than expected. In both localities gas samples were collected at the surface of bubbling streams and may therefore represent a lower-than-equilibrium d 18 O CO2 value.
Kinetic fractionation on bubbles has been extensively studied in relation to the volatile content in degassing volcanic melts (e.g. Aubaud et al., 2004;Paonita and Martelli, 2006). The effect has been observed as d 13 C deviations of dissolved inorganic carbon (DIC) from equilibrium by up to 4‰ in groundwater springs, seepage waters and headwater catchments (Doctor et al., 2008), and up to 2.5‰ deviation from equilibrium in DIC d 13 C samples collected at cold water springs in Green River, Utah (Assayag et al., 2009). The extent of kinetic fractionation increases with water pH and decreases with the depth of degassing. Liquids with lower volatile supersaturation are reported to show lower kinetic fractionation effects as equilibrium can be re-established quicker (Affek and Zaarur, 2014).
The limited availability of d 18 O CO2 measurements does not allow a quantitative comparison between the deviation from equilibrium in springs relative to measured pH and DIC contents. Generally, more actively degassing springs would be expected to deviate from equilibrium more, which agrees with field observations of higher gas discharge rates at Pah Tempe relative to Daylesford.

Summary
Water and CO 2 sampled at the surface is unlikely to be in equilibrium due to the secondary effects of localised partial equilibration and kinetic fractionation during gas ascent to the surface.
Water salinity is unlikely to have an effect. Values of d 18 O CO2 from high flux springs in Pah Tempe and Taradale spring in Daylesford suggest that the oxygen isotope signature from localised equilibration at depth is preserved. CO 2 from diffuse Daylesford springs (e.g. Locarno) re-equilibrates with dilute shallow water. The d 18 O CO2 values of gas collected at the water surface may be further obscured by 18 O depletion during bubble formation, which affects high flux springs more than those with a diffuse low flux.

Implications for usage of d 18 O values in geothermometry
Water 18 O enrichment relative to the MWL with no effect on d 2 H has been traditionally associated with geothermal systems. Waters enriched in 18 O are produced by isotopic exchange between hydrothermal fluids and bedrock minerals, normally at temperatures above 250 C (Clark and Fritz, 1997;D'Amore and Panichi, 1987). The fractionation factor between any mineral and fluid is governed by temperature, thus allowing the distribution of isotopes to be used as a geothermometer (e.g. Giggenbach, 1992). Another mechanism is water-steam separation above liquid-vapour isotopic exchange at 220 C. At this temperature oxygen is kinetically fractionated between fluid and vapour but there is no fractionation in hydrogen isotopes (Clark and Fritz, 1997). Due to the lack of other reported water 18 O enriching mechanisms, it has become common practice to interpret 18 O enrichment in reservoir water as evidence for geothermal conditions (e.g. Cer on et al., 1998;Nelson et al., 2009).
Here, we present evidence that oxygen isotope exchange with CO 2 can result in 18 O-enriched waters, if the starting d 18 O value of CO 2 is significantly higher than that of the water and if high gas to water ratios are present. The oxygen isotope exchange between CO 2 and the spring waters provides a more robust explanation for the 18 O-enriched waters from Pah Tempe springs, which is in closer agreement with the geothermometric calculations and water discharge temperatures without invoking circulation depths of over 5 km and equilibration at temperatures >150 C as quoted in Our findings have two significant implications. Firstly, CO 2water equilibration alone, without the need to invoke any additional processes, may result in water 18 O enrichment or depletion, which means that 18 O-enriched waters should not be solely interpreted as geothermal as is the current practice. Secondly, equilibrium achieved between water and minerals at depth may be obscured by later interaction with CO 2 . Both of these scenarios have significant implications to the sulfate-water oxygen isotope geothermometry technique, which relies on temperature and pHdependant oxygen isotope exchange between water and dissolved SO 4 , applicable to temperature ranges between 140 and 350 C (McKenzie and Truesdell, 1977). The method requires estimation of the water d 18 O in equilibrium with sampled sulfate under an assumption that the original value has not been altered by secondary processes such as dilution with shallow water, boiling and steam loss, near-surface oxidation of H 2 S and biological activity or an application of an appropriate correction (Fowler et al., 2013). Equilibration with CO 2 , which can be achieved in a matter of hours and either deplete or enrich water in 18 O, should also be considered when using this geothermometry technique in CO 2 -rich waters. This consideration may also be significant to palaeowater studies which relate d 18 O values of precipitating phases to either the palaeowater composition or precipitation temperatures (e.g. Astin and Scotchman, 1988;Morad and Eshete, 1990).
A recent study by Ladd and Ryan (2016) demonstrated that shallow surface build up in CO 2 partial pressure and subsequent bubble formation may be the main driving mechanism for geyser eruption in sub-boiling conditions, challenging the common notion that subsurface water boiling is required for this phenomena. Our study provides additional evidence that elevated CO 2 concentrations at ambient temperatures may explain the features often attributed to geothermal systems.

Conclusions
Global natural CO 2 -rich mineral waters show d 18 O deviations from the MWL with no observed change in d 2 H. Oxygen isotope deviations without a change in hydrogen isotopes may be the result of oxygen isotope equilibrium exchange between CO 2 and water, mineral dissolution and re-precipitation, or isotopic exchange with minerals. We have developed a simple geochemical modelling approach to study the influence of low temperature water-rock reactions on oxygen isotope changes in subsurface waters. The method requires knowledge of the water geochemistry (major ion concentrations, dissolved carbon content, pH, temperature) and a conceptual model of reactive and precipitating phases. Numerical modelling can be applied to assess the water-rock interaction influence on oxygen isotope ratios in other saline natural waters or CO 2 storage sites where oxygen isotopes are used as natural tracer of the injected CO 2 plume.
In two case studies from Daylesford (Australia) and Pah Tempe (Utah, USA), we apply our new modelling approach to show that low temperature water-rock reactions are unlikely to have a significant effect on water d 18 O values. In both cases, the water d 18 O shift can be explained by oxygen isotope exchange with CO 2 . Oxygen isotope values observed in the waters measured at Daylesford and Pah Tempe springs are close to equilibrium with d 18 O of the erupting CO 2 . Deviation from ideal equilibrium is likely due to localised CO 2 movement through the water and the establishment of partial equilibration or kinetic isotope fractionation on degassing bubbles sampled at the water surface.
Traditionally, enrichment in 18 O in the reservoir waters relative to the MWL has been interpreted to be the result of geothermal activity, while 18 O depletion is proposed to be due to CO 2 -water interaction at lower temperatures. Our global dataset of oxygen and hydrogen isotope measurements in waters from low temperature CO 2 springs and the case studies presented from the Daylesford and Pah Tempe CO 2 springs provide evidence that equilibration with CO 2 can result in both 18 O enrichment and depletion in spring waters and therefore geothermal conditions are not necessary to produce 18 O-enriched waters. This should be considered in future studies, and used to re-interpret data from previous studies using the water and mineral stable isotope composition to infer water circulation depths, temperatures and local tectonic settings.

R.
Karolyt _ e acknowledges the support of an EPSRC PhD studentship in partnership with Badley Geoscience Ltd. The authors acknowledge the funding provided by CO2CRC (036610/1). S. Serno was funded by the UK Carbon Capture and Storage Research Centre (UKCCSRC) Call 2 grant, G. Johnson and S. Gilfillan were partially supported by both UKCCSRC and Scottish Carbon Capture and Storage (SCCS) (EP/K000446/1). Steve Nelson of Brigham Young University, Utah is thanked for introducing Gilfillan to the Pah Tempe springs and assisting with gas sample collection and background data. We thank Allan Chivas for his assistance in the field during collection of samples for the Daylesford springs and obtaining measurements of water samples. We extend our thanks to Ian Cartwright for sharing his extensive knowledge on Daylesford springs and Jen Roberts for helping out with the fieldwork in Daylesford. We thank Terry Donnelly, Adrian Boyce and Tony Fallick at SUERC for their assistance in obtaining stable isotope measurements of gas samples.