Geothermometry and Isotope Geochemistry of CO 2 -Rich Thermal Waters in Choygan, East Tuva, Russia

: The Choygan area of southern Siberia, Russia hosts a variety of CO 2 -rich thermal mineral and other waters emerging from springs at temperatures between 7 ◦ C and 39 ◦ C. Chemical analyses of the spring waters ( n = 33) were carried out to characterise the waters and determine their origin. A continuum of compositions was observed between relatively lower temperature (7 ◦ C) HCO 3 -Ca-Na dominated waters with relatively low amount of total dissolved solids (TDS) and high Eh, and higher temperature (39 ◦ C) HCO 3 -Na-Ca dominated waters with higher TDS and lower Eh—this reﬂects largely conservative mixing of these components between near surface low temperature, oxidising groundwaters and higher temperature, more reducing thermal waters derived from a deeper geothermal reservoir. Stable isotopic data are consistent with all the water ultimately being derived from meteoric water that has undergone varying degrees of isotopic fractionation following evaporation. The inferred δ 18 O and δ 2 H isotopic composition of the unfractionationed meteoric waters is lighter than that expected that of mean annual local precipitation, which together with a strong negative correlation between δ 18 O and the elevation of the sampled discharging springs, suggests recharge at higher elevations (1600 m to 3000 m; average 2600 m). Reservoir temperature, calculated using geothermometers and an analysis of saturation indices of plausible reservoir minerals, ranged from 70 ◦ C to 100 ◦ C at an inferred depth of 2 to 3 km. Not all chemical components were found to follow conservative mixing behaviour. In particular, (i) the CO 2 contents of the waters were highly variable, suggesting either varying degrees of degassing and/or near discharge admixture with air, and (ii) SO 4 concentrations in the lower temperature thermal CO 2 -rich waters were highly variable, suggesting a role of near surface oxidation processes, for example of pyrite, in modifying the concentration of redox sensitive components. Limited δ 13 C data are consistent with the CO 2 predominately being derived from dissolution of metamorphic/igneous carbonate minerals in the reservoir. Based on geological conditions, isotope and chemical data, a conceptual circulation model of the Choygan hydrothermal system is proposed. stage


Introduction
One current challenge in geochemistry is determining the genesis and particularly the geochemical processes controlling the chemical compositions of mineral waters, which are often generated in complex hydrogeological systems. These waters, particularly carbon dioxide-rich mineral waters, are widely used for a variety of purposes including geothermal power, carbon dioxide production, heating of shelters and greenhouses as well as bottled water and spas. Better understanding their origins may be helpful to inform sustainable development of these resources.
The Choygan mineral water spa, used for several decades by locals and tourists for bathing and medicinal purposes, is an example of such an occurrence of carbon dioxide-rich mineral waters. Located in the Eastern Sayan Mountains, Tuva, Russia, some of these springs are characterized by the bubbling of CO 2 -rich gases. Our group [26,27] have determined the dissolved gas [26], 222 Rn [26] and major and trace chemical compositions of these waters [28,29] (see also previous work summarized in [26,27] as well as making an assessment of the extent of thermodynamic equilibrium of these waters with various minerals [26]). These have enabled preliminary estimates of reservoir temperatures [30,31]. Further constraints on the origins of these waters have been provided by 3 He/ 4 He based heat flow estimates [32]. Notwithstanding these studies, the difficulty of access has precluded more detailed studies of these mineral waters, in particular in relation to geothermometry, isotope geochemistry and the application of these techniques to developing a more detailed conceptual model of the origins of the mineral waters.
This study aims to conduct a more detailed survey of the Choygan CO 2 -rich thermal waters and to better define the geological and geochemical conditions of formation of the mineral waters and, in particular, the extent and nature of water-rock reaction processes (cf. [1, 3,33]). We apply hydrogeochemical geothermometers [4,9,22] to establish the maximum reservoir temperatures, ascertain the effect of CO 2 and temperature on the main processes that control the water chemistry and to estimate the depth of the geothermal system (cf. [4,9,22]). We use stable isotopes (δ 18 O and δ 2 H) to provenance the waters (cf., [1,26]) and to better understand underground flow paths. Lastly, this study aims to provide an improved conceptual circulation model of the Choygan hydrothermal system.

Study Area
The Choygan mineral water natural spa is located in the northeastern part of the Tuva Republic (Russia) at a height of around 1550 m and in an area characterized by a continental climate, a mean annual temperature −4.7 • C with an annual average precipitation of 700 mm (Figure 1) [34]. The groundwaters discharge at 33 local springs within a relatively small area on the western slope of the Eastern Sayan Mountains (Figure 2). The springs are predominantly located on the right bank of the Arzhaan-Hem River, on the first river terraces along the slope as well as in the floodplain of the left riverbank.The combined total discharge from all of the springs is estimated to be around 18 L/s. The Choygan springs lie at the boundary of the Riphean Tuva-Mongolian massif with older Caledonian structures of the Eastern Tuva. The springs are located at the intersection of the topographic low due to the Arzhan-Hem river and the zone of the deep E-W trending Azassko-Zhomboloksky fault. The springs emerge in Lower Paleozoic intrusives on the right bank of the Arzhan-Hem River and crystalline schists, gneisses and marbles on the left bank ( Figure 1). The Eastern Sayan Mountains, in which the Choygan springs are located, are characterized by high seismic activity and young volcanoes. A small field of the Cenozoic basalts is situated 20 km northeast of the Choygan springs.   4, 5, 6, 7, 8, 9, 9a, 11, 12, 13; black dots-group of samples 16, 17, 19, 20, 21a, 22, 23, 24, 25, 26, 27, 28, 29, 29a, 30, 31, 32; white dash-group of samples 1, 2, 3, 10, 15; black dash-spring 33.

Materials and Methods
Water samples (n = 33) including those from both higher temperature (T > 23 °C) and lower temperature (T < 23 °C) springs were collected in July 2013. Samples for analysis for ion chromatography or titrimetric analysis were collected in 500-mL clean polyethylene bottles after rinsing twice with the sampled water. The samples for trace element analyses were acidified with concentrated HNO3 to prevent metal precipitation and stored in 50-mL clean polyethylene bottles. Spring temperature, pH, Eh and electric conductivity (EC) were measured in situ using a portable water test kit (Hanna Instruments, Vöhringen, Germany). In addition, dissolved CO2 was measured in the field by titration with NaOH.
The concentration of HCO3 − was estimated via titration with a 0.01 M solution of HCl using a methyl orange indicator. Ions, including SO4 2− , Cl − , Ca 2+ , Mg 2+ , Na + and K + , were analysed using ion chromatography (Dionex 1000 and Dionex 2000). Dissolved trace elements were determined by inductively coupled mass spectrometry (Perkin Elmer Inc., ELAN-DRC-e, Shelton, CT, USA). All standard solutions for these analyses were prepared with ultra-pure deionized water and Perkin Elmer Multi Element Standard Solutions. Ionic charge balance was calculated and used as a quality control check. The value of the computed ionic charge balance was within an acceptable limit of ±5% for most samples.
The radon content of the water samples was determined using a radiometer RGA-01 (Allbiz, Zelenograd, Russia), which registered the volume alpha-radiation activity of the radon-222 nuclide in liquid samples.

Materials and Methods
Water samples (n = 33) including those from both higher temperature (T > 23 • C) and lower temperature (T < 23 • C) springs were collected in July 2013. Samples for analysis for ion chromatography or titrimetric analysis were collected in 500-mL clean polyethylene bottles after rinsing twice with the sampled water. The samples for trace element analyses were acidified with concentrated HNO 3 to prevent metal precipitation and stored in 50-mL clean polyethylene bottles. Spring temperature, pH, Eh and electric conductivity (EC) were measured in situ using a portable water test kit (Hanna Instruments, Vöhringen, Germany). In addition, dissolved CO 2 was measured in the field by titration with NaOH.
The concentration of HCO 3 − was estimated via titration with a 0.01 M solution of HCl using a methyl orange indicator. Ions, including SO 4 2− , Cl − , Ca 2+ , Mg 2+ , Na + and K + , were analysed using ion chromatography (Dionex 1000 and Dionex 2000). Dissolved trace elements were determined by inductively coupled mass spectrometry (Perkin Elmer Inc., ELAN-DRC-e, Shelton, CT, USA). All standard solutions for these analyses were prepared with ultra-pure deionized water and Perkin Elmer Multi Element Standard Solutions. Ionic charge balance was calculated and used as a quality control check. The value of the computed ionic charge balance was within an acceptable limit of ±5% for most samples. The radon content of the water samples was determined using a radiometer RGA-01 (Allbiz, Zelenograd, Russia), which registered the volume alpha-radiation activity of the radon-222 nuclide in liquid samples.
The analysis of the chemical and gas compositions of the waters was performed at the Laboratory of Hydrogeochemistry of the "Voda" (Water) Research Centre of the Institute of Natural Resources at Tomsk Polytechnic University.
Water δ 18 O and δ 2 H (n = 19) were analysed at the Resource Centre 'Geomodel' Research Park of Saint-Petersburg State University (St. Petersburg, Russia) on a Picarro L-2120i laser spectrometer (Picarro Inc., Santa Clara, CA, USA). V-SMOW-2, GISP, SLAP, USGS45 and USGS46 were used as standards. Accuracy is estimated to be better than ±0.1% for δ 18 O and ±1% for δ 2 H. A limited number of further samples were analysed for δ 13 C at the Laboratory of Isotope and Analytical geochemistry (Novosibirsk, Russia).
Silica (quartz and chalcedony) and cation (Na-K-Ca and Na-K) geothermometers were used to estimate the reservoir temperatures of the Choygan geothermal system using the chemical analyses of the thermal waters (Table 1) using the equations as follows: Quartz geothermometer (Fournier and Potter, 1982) [37]:  (6) where T is the model reservoir temperature in • C; S is the silica concentration of the spring discharge in mg/L; and Na, K and Ca concentrations are expressed in mol/L [37]. The depth of origin of the thermal waters was calculated using the following equation: where T is the model reservoir temperature in • C, T s is the annual mean surface temperature in • C, and γ is thereported regional geothermal gradient in • C/km. In the absence of reservoir rock samples, mineral equilibrium calculations, through geothermometry or calculation of saturation indices (SI), are important methods to assess likely reactive minerals in the subsurface from groundwater chemical data [3]. Mineral saturation indices (SI) for some key carbonate (calcite, aragonite and dolomite), silicate (albite, muscovite and kaolinite), sulphate (gypsum and anhydrite) and silica (quartz and chalcedony) minerals were used to predict the tendency for precipitation or dissolution of those minerals with SI = 0 indicating thermodynamic equilibrium; SI > 0 indicating oversaturation, and SI < 0 indicating undersaturation. Mineral saturation indices of waters from the study area were calculated at the in situ measured spring temperatures and pH using the computer program PHREEQC (version 3, USGS, Denver, CO, USA). The Geochemist's Workbench program (GWB) [38] (Aqueous Solutions LLC, Alexandria, VA, USA) was used to plot activity diagrams.
Following the method of Reed and Spycher (1984) [39], the congruence of mineral saturation indices of potential reservoir minerals over a range of pCO 2 (calculated using Henry's Law and the CO 2 concentrations in the water) and temperatures was used to better estimate reservoir temperatures. The median (RMED), mean (MEAN), standard deviation (SDEV) and root mean-square error (RMSE) of the saturation indices of 9 selected potential reservoir minerals were determined for temperatures between 10 • C and 200 • C and the minimum values of these parameters then used to infer a model most probable reservoir temperature. In order to avoid biases arising from using the saturation indices of minerals with slow reaction kinetics, the median (RMED) was the preferred statistical parameter used to compute reservoir temperature, with SDEV or RMSE used to independently estimate the quality of clustering of saturation indices [39,40].

Hydrogeochemical Characteristics of the Springs
The major ion chemical compositions of the spring water samples are presented in Table 1 and plotted in a Piper diagram ( Figure 3). The spring waters have been grouped into (I) "Higher" temperature (T > 23 • C) thermal CO 2 -rich waters; (II) "Lower" temperature (T < 23 • C) thermal CO 2 -rich waters; and (III) shallow (non-thermal) groundwaters. Data from our group [26] for samples previously collected from the same area in 2011 are also included for comparison. Chemical compositional difference between waters sampled in 2011 [26] and 2013 (this study) were generally insignificant or small, with observed differences in TDS of typically just 10-200 mg/L, fluorine (0.1-0.9 mg/L) and iron (0.1-1 mg/L); however, larger differences in TDS (400-500 mg/L) were observed in several springs, perhaps reflecting changes in weather and hydrological conditions impacting on mixing between groundwaters of whatever type and rainwater [41].
The first group of 18 relatively high temperature (22 • C to 39 • C) thermal springs is located in the central part of the river floodplain, including that on the left river bank. These waters are characterized by low Eh (−170 V to 142 mV). These "higher temperature" thermal CO 2 -rich waters are HCO 3 -Na-Ca type, slightly acidic (pH 6.1-6.9) waters with TDS from 1545 mg/L to 2647 mg/L. The TDS values and concentrations of major ions such as HCO 3 − , Cl − , Ca 2+ , Na + , Mg 2+ , K + and Si 4+ increase with measured water temperature and are higher than those in the colder springs. This likely reflects, in part, longer circulation and residence times of the higher temperature water, with HCO 3 ultimately derived from the dissolution of carbonate minerals and of CO 2 , and Na + , K + , Ca 2+ , and Si 4+ in groundwater associated with the dissolution of silicate phases in the subsurface [3,33]. The coating of a red precipitate around the springs is a result of the oxidation of Fe II to Fe III upon exposure to the atmosphere. This iron may have been leached from biotite and pyroxenes contained in gneisses and crystalline schists in the subsurface. The second group of 12 relatively low temperature thermal CO2-rich waters found in the northeast and southwest parts of the study area are of the НСО3-Ca-Na type.These waters are slightly acidic (pH 5.9 to 6.7) and with relatively high Eh (170 mV to 236 mV). In contrast to the first group, the TDS values of the second group are relatively low, ranging from 607 to 2064 mg/L, whereas the CO2 concentration reaches 1488 mg/L with an average of 817 mg/L. The higher CO2 values may reflect, in part,the higher solubility of CO2 at lower temperatures. The low Fe (0.1-0.2 mg/L) compared to the "higher temperature" thermal waters (0.1-4.4 mg/L) and relatively high concentration of SO4 2− (5-59 mg/L) in the low temperature water are both consistent with the relatively oxidizing Eh conditions in this group of springs. K and Mg are found in low concentrations in both the higher and lower temperature thermal waters corresponding to their low content in the metamorphic bedrocks.
The discharge of shallow groundwater is represented by three springs situated on the left riverbank and on the slopes of the right bank. This group of НСО3-Са type waters shows that the lowest discharge temperatures (7 °C to 14 °C) is circumneutral/weakly alkaline (pH 6.6-8.3) with relatively high Eh (169 mV to 224 mV) and low TDS (290 to 350 mg/L), these low values being consistent with low levels of water-rock interactions. The observed discharge temperatures of these Group III waters are consistent, assuming a plausible local geothermal gradient of 33 °C/km [32], with conductive heating at depths of between 0 and 400 m of waters of meteoric origin and with recharge temperatures between 0 °C and average daily maximum air temperatures of around 13 °C. Notwithstanding this, some mixing with thermal waters cannot be discounted.
The relationship between various chemical constituents and the conservative tracer [1,42], Cl − , in the three groups of waters are presented in Figure 4. The strong positive correlations between all of Na + , Mg 2+ , K + and HCO3 − with Cl − , is broadly consistent with the major element chemistry being controlled by mixing between a higher temperature thermal CO2-rich water (cf. Group I) and lower temperature shallow groundwater (cf. Group III)-indeed, with the notably exception of SO4 2− , the major element compositions of the lower temperature thermal CO2-rich waters (Group II) are generally intermediate between those of Group I and Group III, and those of all the waters generally intermediate between the compositions of Spring 10 and Spring 33, which represent the observed extreme compositional end-members of Group I and Group III respectively. Notwithstanding this, Ca 2+ is somewhat less correlated with Cl − , suggesting that its variation is also controlled to some degree by variable dissolution of carbonate minerals and degassing of CO2, which in turn controls HCO3 − concentrations. Several samples also exhibit relatively elevated SO4 2− due to a different process, possibly the oxidation of iron sulphide minerals, such as pyrite. The second group of 12 relatively low temperature thermal CO 2 -rich waters found in the northeast and southwest parts of the study area are of the HCO 3 -Ca-Na type.These waters are slightly acidic (pH 5.9 to 6.7) and with relatively high Eh (170 mV to 236 mV). In contrast to the first group, the TDS values of the second group are relatively low, ranging from 607 to 2064 mg/L, whereas the CO 2 concentration reaches 1488 mg/L with an average of 817 mg/L. The higher CO 2 values may reflect, in part, the higher solubility of CO 2 at lower temperatures. The low Fe (0.1-0.2 mg/L) compared to the "higher temperature" thermal waters (0.1-4.4 mg/L) and relatively high concentration of SO 4 2− (5-59 mg/L) in the low temperature water are both consistent with the relatively oxidizing Eh conditions in this group of springs. K and Mg are found in low concentrations in both the higher and lower temperature thermal waters corresponding to their low content in the metamorphic bedrocks. The discharge of shallow groundwater is represented by three springs situated on the left riverbank and on the slopes of the right bank. This group of HCO 3 -Ca type waters shows that the lowest discharge temperatures (7 • C to 14 • C) is circumneutral/weakly alkaline (pH 6.6-8.3) with relatively high Eh (169 mV to 224 mV) and low TDS (290 to 350 mg/L), these low values being consistent with low levels of water-rock interactions. The observed discharge temperatures of these Group III waters are consistent, assuming a plausible local geothermal gradient of 33 • C/km [32], with conductive heating at depths of between 0 and 400 m of waters of meteoric origin and with recharge temperatures between 0 • C and average daily maximum air temperatures of around 13 • C. Notwithstanding this, some mixing with thermal waters cannot be discounted.
The relationship between various chemical constituents and the conservative tracer [1,42], Cl − , in the three groups of waters are presented in Figure 4. The strong positive correlations between all of Na + , Mg 2+ , K + and HCO 3 − with Cl − , is broadly consistent with the major element chemistry being controlled by mixing between a higher temperature thermal CO 2 -rich water (cf. Group I) and lower temperature shallow groundwater (cf. Group III)-indeed, with the notably exception of SO 4 2− , the major element compositions of the lower temperature thermal CO 2 -rich waters (Group II) are generally intermediate between those of Group I and Group III, and those of all the waters generally intermediate between the compositions of Spring 10 and Spring 33, which represent the observed extreme compositional end-members of Group I and Group III respectively. Notwithstanding this, Ca 2+ is somewhat less correlated with Cl − , suggesting that its variation is also controlled to some degree by variable dissolution of carbonate minerals and degassing of CO 2 , which in turn controls

Dissolved Gas Composition
The gas composition of the studied springs is shown in Table 1and Figure 5 and reveals a relatively constant N 2 :O 2 ratio (4.1 ± 1.0 mol/mol) just marginally higher than that of air (3.7) and with intermediate but highly variable proportions of CO 2 . This suggests that the gases sampled are essentially mixtures of air and CO 2 with minor proportions of excess (i.e., above atmospheric composition) N 2 . The CO 2 content of the analysedranged from 6 vol.% to 65 vol.%, with a broad overlap of compositions between Group I and Group II, but with the CO 2 contents of the shallow groundwaters (Group III) being distinctly lower than that of the thermal waters ( Figures 5 and 6). Whilst relatively high oxygen contents are known to be characteristic of near surface waters [43], for the thermal waters, the variable oxygen (and highly correlated nitrogen) contents might reflect variable admixture of air during sampling of the springs at discharge. Highly variable gas CO 2 contents may also reflect this process and/or CO 2 exsolution/solution processes.

Dissolved Gas Composition
The gas composition of the studied springs is shown in Table 1and Figure 5 and reveals a relatively constant N2:O2 ratio (4.1 ±1.0 mol/mol) just marginally higher than that of air (3.7) and with intermediate but highly variable proportions of CO2. This suggests that the gases sampled are essentially mixtures of air and CO2 with minor proportions of excess (i.e., above atmospheric composition) N2. The СО2 content of the analysedranged from 6 vol.% to 65 vol.%, with a broad overlap of compositions between Group I and Group II, but with the CO2 contents of the shallow groundwaters (Group III) being distinctly lower than that of the thermal waters ( Figures 5 and 6). Whilst relatively high oxygen contents are known to be characteristic of near surface waters [43], for the thermal waters, the variable oxygen (and highly correlated nitrogen) contents might reflect variable admixture of air during sampling of the springs at discharge. Highly variable gas CO2 contents may also reflect this process and/or CO2 exsolution/solution processes.

4.3.Radon-222 Concentration
The CO2-rich Choygan waters contain appreciable radon (Table 1). However, the compositions of the higher temperature and lower temperature thermal CO2-rich waters are broadly similar (4 Bq/L to 948 Bq/L cf. 53 Bq/L to 519 Bq/L) and, although elsewhere, higher radon concentrations are observed in waters with low mineralization [44]; in the Choygan waters, such a relationship is not

Dissolved Gas Composition
The gas composition of the studied springs is shown in Table 1and Figure 5 and reveals a relatively constant N2:O2 ratio (4.1 ±1.0 mol/mol) just marginally higher than that of air (3.7) and with intermediate but highly variable proportions of CO2. This suggests that the gases sampled are essentially mixtures of air and CO2 with minor proportions of excess (i.e., above atmospheric composition) N2. The СО2 content of the analysedranged from 6 vol.% to 65 vol.%, with a broad overlap of compositions between Group I and Group II, but with the CO2 contents of the shallow groundwaters (Group III) being distinctly lower than that of the thermal waters ( Figures 5 and 6). Whilst relatively high oxygen contents are known to be characteristic of near surface waters [43], for the thermal waters, the variable oxygen (and highly correlated nitrogen) contents might reflect variable admixture of air during sampling of the springs at discharge. Highly variable gas CO2 contents may also reflect this process and/or CO2 exsolution/solution processes.

4.3.Radon-222 Concentration
The CO2-rich Choygan waters contain appreciable radon (Table 1). However, the compositions of the higher temperature and lower temperature thermal CO2-rich waters are broadly similar (4 Bq/L to 948 Bq/L cf. 53 Bq/L to 519 Bq/L) and, although elsewhere, higher radon concentrations are observed in waters with low mineralization [44]; in the Choygan waters, such a relationship is not

Radon-222 Concentration
The CO 2 -rich Choygan waters contain appreciable radon (Table 1). However, the compositions of the higher temperature and lower temperature thermal CO 2 -rich waters are broadly similar (4 Bq/L to 948 Bq/L cf. 53 Bq/L to 519 Bq/L) and, although elsewhere, higher radon concentrations are observed in waters with low mineralization [44]; in the Choygan waters, such a relationship is not apparent, there also being no significant relationship between radon and Na + , Ca 2+ and Mg 2+ . The short half-life (3.82 days [45]) of radon-222, means that, in contrast to carbon dioxide, the source of radon is likely located near the springs, most probably from felsic rocks, enriched in uranium, and from which radon is produced by the alpha decay of radium-226, which is part of the radioactive uranium-238 series [44].

Geothermometers
Model reservoir temperatures, estimated using various geothermometers from the measured chemical compositions of the spring waters, are presented in Table 2. The lowest estimated reservoir temperatures are found using the chalcedony geothermometer, viz. 42-77 • C for the Group I (higher temperature thermal waters) and 31-52 • C forthe Group II (lower temperature thermal waters). The temperatures found by the quartz geothermometers are somewhat higher, viz. 74-107 • C for the higher temperature thermal water and 61-84 • C for the lower temperature thermal waters. The Na-K-Ca geothermometer gave rise to intermediate temperatures between the Na-K and silica geothermometers, viz. 84-119 • C and 42-84 • C, for the Group I and Group II thermal waters respectively. Significantly higher computed deep temperatures are obtained when using different Na-K geothermometers for low temperature thermal water as opposed to higher temperature thermal waters due to the high Ca 2+ and Na + concentrations. The Na-K geothermometers give similar temperature values from 210 • C to 285 • C for both high and low temperature thermal CO 2 -rich waters. Those temperatures are in disagreement with the values obtained with Na-K-Ca and silica geothermometers.
Mixing of thermal waters and colder groundwaters as well as a lack of a full chemical equilibrium may cause chemical geothermometers to provide uncertain results [3]. One of the requirements for the successful application of cation geothermometers is the attainment of water-rock equilibrium in the geothermal reservoir; the equilibrium can be evaluated by the Na-K-Mg triangular diagram [6]. The groundwater is evaluated with a Na-K-Mg ternary diagram proposed by Giggenbach (1988) [46] to identify the degree of maturation. According to the Na-K-Mg ternary diagram (Figure 7), all samples fall within the immature field, in particular, the Group III (shallow groundwaters) are very immature, consistent with unrealistically high model temperatures obtained by this approach, and both groups of thermal CO 2 -rich waters are not in equilibrium, most likely because of being mixed with colder waters. Therefore, cation geothermometers do not likely yield realistic equilibration temperatures for these waters. The Na-K and Na-K-Ca geothermometers appear to provide unreliable estimations of the reservoir temperatures, which are too high (up to 285 • C). The quartz geothermometer is more suitable than cation geothermometers for such immature waters, giving lower and more reliable temperatures (63-107 • C) for the Choygan geothermal system.

Mineral Saturation States (SI)
As shown in Table 3, the waters of all the studied springs are oversaturated with respect to kaolinite, muscovite, k-feldspar and far from saturation with respect to anhydrite and gypsum at their discharge temperatures. However, in samples that have high SO 4 (>0.4 meq/L),the degree of undersaturation of sulfate minerals is lower and some springs are even in equilibrium with these minerals. The saturation indices of the silica minerals are above zero for both higher and lower temperature thermal CO 2 -rich waters. Higher temperature thermal CO 2 -rich waters are oversaturated or nearly in equilibrium with respect to the carbonate minerals and albite, while these minerals are unsaturated in the lower temperature thermal waters. The mineral saturation indices indicate the equilibrium state between higher temperature thermal waters and minerals such as calcite, aragonite and dolomite at the discharge temperatures and reflect a long residence time. Such behaviour is predictable due to the loss of CO 2 from the solution at atmospheric pressure, and, as a consequence, increases the solution pH, resulting in calcite precipitation in the discharge area. The shallow groundwater is undersaturated with respect to albite, kaolinite, muscovite, k-feldspar, silica and sulfate minerals and oversaturated with respect to carbonate minerals, reflecting the initial stage of water-rock interaction. Of course, thermal water compositions are more likely to reflect equilibrium between the waters and minerals (i.e., saturation) at deeper reservoir temperatures rather than at the discharge temperature. Thus, plotting SI with respect to several hydrothermal minerals as a function of temperature for several higher temperature thermal CO 2 -rich springs (Figure 8) enables the intersection of the equilibrium lines for a group of selected minerals close to SI = 0 to be interpreted to reflect the most likely reservoir temperatures of those waters [3]. Of course, thermal water compositions are more likely to reflect equilibrium between the waters and minerals (i.e., saturation) at deeper reservoir temperatures rather than at the discharge temperature. Thus, plotting SI with respect to several hydrothermal minerals as a function of temperature for several higher temperature thermal CO2-rich springs (Figure 8) enables the intersection of the equilibrium lines for a group of selected minerals close to SI = 0 to be interpreted to reflect the most likely reservoir temperatures of those waters [3]. Whilst calcite, dolomite, aragonite, albite and quartz minerals are oversaturated or in equilibrium at the outflow conditions at 30-40 °C, which does not reflect the reservoir temperature, the SIs of muscovite, kaolinite, magnesite, chalcedony and quartz for the higher temperature thermal CO2-rich waters converge and lie close to a SI = 0 at around 90-100 °C. This is the temperature at which the maximum number of mineral phases are in equilibrium with the waters and is thus interpreted as corresponding to the reservoir temperature. Anhydrite appears to be undersaturated at all temperatures and is of limited value in assessing likely reservoir temperatures. Calcite, aragonite, dolomite and magnesite are oversaturated at the modelled reservoir temperatures but closer to equilibrium at the observed discharge temperatures-this may reflect their involvement in lower temperature processes and/or the influence of changeable pCO2-either way consideration of the temperature dependence of their SIs does not provide a reliable estimate of reservoir temperature.
The aqueous activities of Ca ++ (aq), Mg ++ (aq), Na + (aq), K + (aq), and H + (aq) for the Choygan springs are plotted in M z+ (aq)/aH +( aq) z vs. SiO2 activity-activity diagrams at temperatures of 25 °C, 100 °C, and 170 °C, bracketing the likely ranges of reservoir and near surface temperatures of the thermal waters and groundwaters (Figure 9). Whilst calcite, dolomite, aragonite, albite and quartz minerals are oversaturated or in equilibrium at the outflow conditions at 30-40 • C, which does not reflect the reservoir temperature, the SIs of muscovite, kaolinite, magnesite, chalcedony and quartz for the higher temperature thermal CO 2 -rich waters converge and lie close to a SI = 0 at around 90-100 • C. This is the temperature at which the maximum number of mineral phases are in equilibrium with the waters and is thus interpreted as corresponding to the reservoir temperature. Anhydrite appears to be undersaturated at all temperatures and is of limited value in assessing likely reservoir temperatures. Calcite, aragonite, dolomite and magnesite are oversaturated at the modelled reservoir temperatures but closer to equilibrium at the observed discharge temperatures-this may reflect their involvement in lower temperature processes and/or the influence of changeable pCO 2 -either way consideration of the temperature dependence of their SIs does not provide a reliable estimate of reservoir temperature.
The aqueous activities of Ca ++ (aq), Mg ++ (aq), Na + (aq), K + (aq), and H + (aq) for the Choygan springs are plotted in M z+ (aq)/aH +( aq) z vs. SiO 2 activity-activity diagrams at temperatures of 25 • C, 100 • C, and 170 • C, bracketing the likely ranges of reservoir and near surface temperatures of the thermal waters and groundwaters (Figure 9). Two groups are clearly distinguishable: Group 1 includes nine lower temperature and four higher temperature thermal CO2-rich springs and Group 2 consisting of 14 higher temperature and Two groups are clearly distinguishable: Group 1 includes nine lower temperature and four higher temperature thermal CO 2 -rich springs and Group 2 consisting of 14 higher temperature and one lower temperature thermal CO 2 -rich springs. Interestingly, the springs of Group 1 are located on the periphery of the study area, whereas the springs of Group 2 are found in the central part of the study area.
Group 1 waters fall into the kaolinite stability field at 25 • C, but more closely reflect the coexistence of gibbsite, kaolinite, muscovite and Ca-, Mg-, K-montmorillonite phases at 100 • C. Similarly, Group 2 water compositions most closely reflect coexistence of multiple silicate phases at 100 • C (cf. 25 • C or 170 • C) consistent with saturation states calculated from the chemical geothermometers. Thus, to a large exent, the mineral assemblage kaolinite-muscovite-Ca-, Mg-, Na-, and K-montmorillonite influences the chemical composition of both the higher and lower temperature thermal CO 2 -rich waters, whilst 100 • C appears to be a highly plausible best reservoir temperature of the Choygan geothermal system.
In addition to temperature, the partial pressure of CO 2 affects the precipitation and dissolution of different minerals. The solubility of CO 2 in water decreases with increasing temperature and salinity but increases with pressure [47]. Calculations of saturation indices were therefore performed at the inferred reservoir temperature (95 • C) and at different CO 2 pressures to demonstrate the influence of pCO 2 (Figure 10), viz. as pCO 2 increases the pH values and carbonate and iron mineral saturation indices states rapidly decrease, but the SI values of the silicate minerals are largely unaffected.
Water 2017, 9, x FOR PEER REVIEW 18 of 29 one lower temperature thermal CO2-rich springs. Interestingly, the springs of Group 1 are located on the periphery of the study area, whereas the springs of Group 2 are found in the central part of the study area. Group 1 waters fall into the kaolinite stability field at 25 °C, but more closely reflect the coexistence of gibbsite, kaolinite, muscovite and Ca-, Mg-, K-montmorillonite phases at 100 °C. Similarly, Group 2 water compositions most closely reflect coexistence of multiple silicate phases at 100 °C (cf. 25 °C or 170 °C) consistent with saturation states calculated from the chemical geothermometers. Thus, to a large exent, the mineral assemblage kaolinite-muscovite-Ca-, Mg-, Na-, and K-montmorillonite influences the chemical composition of both the higher and lower temperature thermal CO2-rich waters, whilst 100°C appears to be a highly plausible best reservoir temperature of the Choygan geothermal system.
In addition to temperature, the partial pressure of CO2 affects the precipitation and dissolution of different minerals. The solubility of CO2 in water decreases with increasing temperature and salinity but increases with pressure [47]. Calculations of saturation indices were therefore performed at the inferred reservoir temperature (95 °C) and at different CO2 pressures to demonstrate the influence of pCO2 (Figure 10), viz. as pCO2 increases the pH values and carbonate and iron mineral saturation indices states rapidly decrease,but the SI values of the silicate minerals are largely unaffected. The multicomponent geothermometry approach for temperature estimations included statistical analyses of saturation indices the following minerals: muscovite, kaolinite, montmorillonite (Ca, Na, K, Mg), quartz, and chalcedony. These minerals were chosen as a potential mineral assemblage of Choygan geothermal system following on results of the saturation indices and equilibrium states and controlling of chemistry of water samples. The results of statistical analyses of saturation indices are shown in Table 4.
The computed mineral saturation indices (log(Q/K)) as a function of temperature indicate reservoir temperatures largely ranging from 40 °C (in Group II) to 80 °C (in Group I) with no convergence established for the shallow groundwater samples and significantly higher temperatures, 130 °C and 180 °C, determined for samples 16 and 19, respectively ( Figure 11; Table 4).
Summarizing the saturation states of the selected minerals (Figures 8 and 9) and the statistical analysis of the saturation indices (Figure 11), the reservoir temperature is estimated to be between 70 °C and 100 °C for the Choygan geothermal system. This is consistent with the quartz geothermometer derived reservoir temperature of 98 °C. This estimated reservoir temperature at a depth of 3 km is consistent with the area's high heat flow (84 MW/m 2 ) and geothermal gradient (33.6 °C/km) [32], determined elsewhere by an isotope helium assessment in the Eastern Tuva territory [32]. Compared to the Eastern Tuva, in the Western Tuva, the average heat flow is much lower, ~45-50 mW/m 2 . The multicomponent geothermometry approach for temperature estimations included statistical analyses of saturation indices the following minerals: muscovite, kaolinite, montmorillonite (Ca, Na, K, Mg), quartz, and chalcedony. These minerals were chosen as a potential mineral assemblage of Choygan geothermal system following on results of the saturation indices and equilibrium states and controlling of chemistry of water samples. The results of statistical analyses of saturation indices are shown in Table 4.
The computed mineral saturation indices (log(Q/K)) as a function of temperature indicate reservoir temperatures largely ranging from 40 • C (in Group II) to 80 • C (in Group I) with no convergence established for the shallow groundwater samples and significantly higher temperatures, 130 • C and 180 • C, determined for samples 16 and 19, respectively ( Figure 11; Table 4).
Summarizing the saturation states of the selected minerals (Figures 8 and 9) and the statistical analysis of the saturation indices (Figure 11), the reservoir temperature is estimated to be between 70 • C and 100 • C for the Choygan geothermal system. This is consistent with the quartz geothermometer derived reservoir temperature of 98 • C. This estimated reservoir temperature at a depth of 3 km is consistent with the area's high heat flow (84 MW/m 2 ) and geothermal gradient (33.6 • C/km) [32], determined elsewhere by an isotope helium assessment in the Eastern Tuva territory [32]. Compared to the Eastern Tuva, in the Western Tuva, the average heat flow is much lower,~45-50 mW/m 2 .      Table 1) provide insight into the origin of the spring waters and processes in the study area. The δ 18 O and δ 2 H values of the higher temperature thermal CO 2 -richwaters range from −14.4% to −18.8% and from −130% to −142%, respectively, those of the lower temperature thermal CO 2 -rich waters range from −13.7% to −18.4% and from −127% to −141%, respectively, and those of the shallow groundwaterwere −16.8% and −133%, respectively. The isotopic composition of local precipitation was not measured but was estimated from the Online Isotope in Precipitation Calculator [48]

Isotopic Composition
Stable isotopes (δ 18 O and δ 2 H, Table 1) provide insight into the origin of the spring waters and processes in the study area. The δ 18 O and δ 2 H values of the higher temperature thermal CO2richwaters range from −14.4% to −18.8% and from −130% to −142%, respectively, those of the lower temperature thermal CO2-rich waters range from −13.7% to −18.4% and from −127% to −141%, respectively, and those of the shallow groundwaterwere −16.8% and −133%, respectively. The isotopic composition of local precipitation was not measured but was estimated from the Online Isotope in Precipitation Calculator [48] to range from δ 18 O = −15.3%, δ 2 H = −117.0% at an elevation of 1601 m to δ 18 O = −18.5%, δ 2 H = −139.0% at an elevation of 3044 m (Figure 12).

Figure 12.
Stable isotopic (δ 18 O, δ 2 H) composition of Group I (higher temperature thermal CO2-rich waters) (red circles), Group II (lower temperature thermal CO2-rich waters) (blue squares) and Group III (shallow groundwater) (green triangles) waters compared to that of modelled atmospheric precipitation (pink circles)-monthly (small symbols) and annual (large symbol) along the indicated local meteoric water line (LMWL, equation on diagram). In the magnified inset, the blue line shows the trend line through all the thermal water samples; the red line that through the groundwater and thermal waters (numbers indicate sample numbers) with the maximum shift in δ 18 O from the LMWL, the slopes of these lines being 2.9 and 1.7, respectively. The isotopic composition of precipitation was calculated using the Online Isotope in Precipitation Calculator [48].
The observed spring water isotopic compositions are broadly consistent with an origin from local precipitation with isotopic enrichment arising from evaporation [49,50]. The intercept of the trend line of the thermal water compositions with the modelled local meteoric water line (LMWL) is slightly more negative than that of modern meteoric water. This might reflect palaeoclimatic changes, or perhaps more likely, particularly given the observed correlation between spring elevation and The observed spring water isotopic compositions are broadly consistent with an origin from local precipitation with isotopic enrichment arising from evaporation [49,50]. The intercept of the trend line of the thermal water compositions with the modelled local meteoric water line (LMWL) is slightly more negative than that of modern meteoric water. This might reflect palaeoclimatic changes, or perhaps more likely, particularly given the observed correlation between spring elevation and δ 18 O, a relatively high elevation (1600 m to 3100 m; average 2600 m; see Figure 13) of the recharge area of part of the thermal water [1] and might, in part, explain the relatively low slope of the trend line of the thermal water compositions-although other processes, such as isotopic exchange arising from water-rock reactions within the geothermal reservoir [26] cannot be discounted.
Water 2017, 9, x FOR PEER REVIEW 24 of 29 δ 18 O, a relatively high elevation (1600 m to 3100 m; average 2600 m; see Figure 13) of the recharge area of part of the thermal water [1] and might, in part, explain the relatively low slope of the trend line of the thermal water compositions-although other processes, such as isotopic exchange arising from water-rock reactions within the geothermal reservoir [26] cannot be discounted. The wide variation in the mineralogy of the host rocks and the P-T conditions of water-rock interaction in the transit area result in a complex modification of the waters' chemical and isotopic compositions. Fissuring of the geological media causes a mixing of the waters that enter the discharge area from recharge areas with different altitudes (and consequently, flow-path distances). δ 13 C data were used to investigate the source of carbon in the groundwater. The δ 13 C (total inorganic dissolved carbon) values observed in the high and low temperature thermal waters samples from springs 5, 10, 11, 13, 15, 18, 19 and 27 were found to range from −0.3 to +1% [51]. Despite the location of springs in a tectonically and seismically active zone with volcanic activity, the presence of large active faults and the presence of mantle helium,the isotopic composition of the CO2 is consistent with metamorphism of carbonate rocks at depth, although inputs from other sources cannot be ruled out.

Conceptual Circulation Model
The geological conditions, chemical, gas and isotopic compositions, as well as the estimations of the reservoir temperatures and depths, of the thermal and cold waters make it possible to propose a conceptual circulation model of the origin, the probable types of underground flow paths and the sources of CO2 in the groundwater of the Choygan natural spa (Figure 14). The appearance of thermal springs is controlled by two factors: the high heat flow in the region (84 MW/m 2 ) and the presence of fracture zones that extend to great depths, allowing hot water and CO2 to quickly rise to the surface. Helium isotope studies of gas and water samples from Choygan (Sample 13) [32] indicate a major (35%) mantle helium component consistent with these elevated heat flows. The wide variation in the mineralogy of the host rocks and the P-T conditions of water-rock interaction in the transit area result in a complex modification of the waters' chemical and isotopic compositions. Fissuring of the geological media causes a mixing of the waters that enter the discharge area from recharge areas with different altitudes (and consequently, flow-path distances). δ 13 C data were used to investigate the source of carbon in the groundwater. The δ 13 C (total inorganic dissolved carbon) values observed in the high and low temperature thermal waters samples from springs 5, 10, 11, 13, 15, 18, 19 and 27 were found to range from −0.3 to +1% [51]. Despite the location of springs in a tectonically and seismically active zone with volcanic activity, the presence of large active faults and the presence of mantle helium, the isotopic composition of the CO 2 is consistent with metamorphism of carbonate rocks at depth, although inputs from other sources cannot be ruled out.

Conceptual Circulation Model
The geological conditions, chemical, gas and isotopic compositions, as well as the estimations of the reservoir temperatures and depths, of the thermal and cold waters make it possible to propose a conceptual circulation model of the origin, the probable types of underground flow paths and the sources of CO 2 in the groundwater of the Choygan natural spa (Figure 14). The appearance of thermal springs is controlled by two factors: the high heat flow in the region (84 MW/m 2 ) and the presence of fracture zones that extend to great depths, allowing hot water and CO 2 to quickly rise to the surface. Helium isotope studies of gas and water samples from Choygan (Sample 13) [32] indicate a major (35%) mantle helium component consistent with these elevated heat flows.
presumably at high altitudes, between about 1600 m and 3000 m (cf. Topographer's Peak, height of 3044 m above sea level, see Figure 1) and percolates, including through faults and other higher permeability structures into the hydrogeological system, where it reacts with primary aluminosilicate minerals that are not in equilibrium with the waters. These reactions result in non-stoichiometric dissolution, particularly silicate minerals inplagiogranites, diorites, schists andgneiss. The water reaches saturation with respect to kaolinite and montmorillonite and carbonate minerals are precipitated, with a concomitant net increase in Na, K, Ca, Mg and TDS. Palaeozoic metamorphics, particularly including highly faulted and fractured marbles and limestones may be the reservoir units supplying CO2 for the Choygan thermal springs. The reservoir temperature of water reaches up to approximately 100 °C at a depth of nearly 3 km. After heating at depth, water rises via major faults toward the surface, accompanied by mixing with cooler, variably more oxidizing groundwaters, and ultimately discharging at the surface, with concomitant precipitation of carbonates (notably travertine), Fe(III)-O-H phases and admixture with air. Radon produced by radioactive decay of Ra-226 originating from uranium-enriched felsic rocks is incorporated into the waters relatively close to the discharge zones.

Conclusions
Thermal CO2-rich mineral waters in the Eastern Sayan Mountainous area in East Tuva (Russia) as exemplified by springs discharging around the Choygan mineral water natural spa, range in composition from higher temperature HCO3-Na-Ca type waters to lower temperature НСО3-Ca-Na type waters and broadly represent mixing between higher Na/Ca geothermal waters and higher The Choygan mineral waters are associated with a large E-W striking fault in the Precambrian rocks (plagiogranites, diorites and granodiorites) intruded by Paleozoic granites and diorites. The isotopic composition of these waters reflects their meteoric origin as well as evaporative processes and perhaps also isotopic exchange during water-rock reactions.
Atmospheric precipitation having ambient surface temperatures infiltrates the ground presumably at high altitudes, between about 1600 m and 3000 m (cf. Topographer's Peak, height of 3044 m above sea level, see Figure 1) and percolates, including through faults and other higher permeability structures into the hydrogeological system, where it reacts with primary aluminosilicate minerals that are not in equilibrium with the waters. These reactions result in non-stoichiometric dissolution, particularly silicate minerals inplagiogranites, diorites, schists andgneiss. The water reaches saturation with respect to kaolinite and montmorillonite and carbonate minerals are precipitated, with a concomitant net increase in Na, K, Ca, Mg and TDS. Palaeozoic metamorphics, particularly including highly faulted and fractured marbles and limestones may be the reservoir units supplying CO 2 for the Choygan thermal springs. The reservoir temperature of water reaches up to approximately 100 • C at a depth of nearly 3 km.
After heating at depth, water rises via major faults toward the surface, accompanied by mixing with cooler, variably more oxidizing groundwaters, and ultimately discharging at the surface, with concomitant precipitation of carbonates (notably travertine), Fe(III)-O-H phases and admixture with air. Radon produced by radioactive decay of Ra-226 originating from uranium-enriched felsic rocks is incorporated into the waters relatively close to the discharge zones.

Conclusions
Thermal CO 2 -rich mineral waters in the Eastern Sayan Mountainous area in East Tuva (Russia) as exemplified by springs discharging around the Choygan mineral water natural spa, range in composition from higher temperature HCO 3 -Na-Ca type waters to lower temperature HCO 3 -Ca-Na type waters and broadly represent mixing between higher Na/Ca geothermal waters and higher Ca/Na near surface groundwaters. All springs emerge along faults, which have the potential to transmit waters rapidly from great depths.
Although mixing appears to be the dominant process controlling major element hydrogeochemistry, the non-conservative behaviour of CO 2 and SO 4 may reflect degassing and/or other processes. All the waters are ultimately of meteoric origin with stable isotopic signatures reflecting recharge from colder environments than that of the modern day discharge areas and isotopic fractionation during evaporative processes. Mixing processes result in water compositions that are strongly out of equilibrium with Na-K-Ca phases; accordingly, the most reliable estimates of geothermal reservoir temperatures are obtained from silica geothermometry and an analysis of temperature-dependent convergence of likely reservoir mineral saturation indices-these suggest reservoir temperatures of around 80 to 100 • C consistent with a previously observed geothermal gradient in the area of around 34 • C/km and a reservoir depth of approximately 3 km.
Future research on determining the isotopic composition of local surface waters (rivers, precipitation, shallow groundwater) as well as more extensive measurements of δ 13 C and δ 18 O are recommended to provide a better understanding of the origin of variable CO 2 concentraions, and in particular to better discriminated between degassing, oxidation and mixing processes in controlling the chemical compositions of these thermal waters.
Author Contributions: The study was designed and conceived by Y.K. and N.G.; A.S. carried out the field work, collected the water and gas samples, undertook data analysis and was the primary author of all sections of the paper, with the exception of the "Isotopic compositions" section, which was primarily written by I.T.; laboratory analyses were carried out by A.K.; D.P. contributed particularly to the development of the sections "Geothermometers" and "Mineral saturation states"; I.T. wrote the section "Isotopic composition". All authors contributed to the writing and review of the final manuscript.