A simple scenario explaining the range of variations of the conductivity and of the chemical compositions of the lake waters would involve an input of concentrated and Na-K-Cl-SO₄-rich waters into fresh waters having geochemical characteristics similar to those of spring waters. The magnitude of the input would vary from lake to lake. Such a binary mixing scenario should however entail that, in mixing diagrams such as for instance X/Cl vs. Na/Cl (X = Ca, Mg, Sr, Ba, Si) or ⁸⁷Sr/⁸⁶Sr vs Cl/Sr diagrams, the data points representative of the lake water samples would plot along a linear mixing trend between a fresh water end-member, represented by the spring waters, and a highly saline end-member represented by the most saline lakes enriched in Na, K, Cl and SO₄. This is clearly not the case (Figs. 2 and 4), therefore ruling out the simple binary mixing scenario. Incidentally, it can be noticed that in all these diagrams the characteristics of the Anga river and Lake Baikal samples are quite similar to those of the spring and ‘diluted lake’ waters, but systematically different of those of the ‘concentrated lake’ waters.

The observation that the lake and spring waters exhibit very similar Sr isotope ratio ranges of variation certainly outlines that the different Sr sources that control the Sr isotope ratios of the waters, such as bedrock minerals or atmospheric deposits, are not really different for lakes and springs. Such a geochemical similarity between lake and spring waters is not only found in Sr isotope ratios. It is also observed in X/Cl element ratios when X is an alkali metal (Na, K or Rb). By contrast it disappears when X is an alkaline earth metal such as Ca, Mg, Sr or Ba (Fig. 4). In this case, the X/Cl ratios are systematically lower in lake waters than in spring waters which could indicate a loss of alkaline earth elements in lake waters compared to spring waters. It might also indicate an input of Cl in lake waters; however such a scenario has already been ruled out (see above). Usually, the first minerals to precipitate during evaporation of natural waters are alkaline earth metal bearing minerals, whereas alkali metal bearing minerals only precipitate when water has reached a sufficiently high degree of concentration (Jones and Deocampo, 2005). For this reason, and on the basis of our data, we propose that the range of variations of chemical composition and conductivity of the lake waters results from a water evaporation or water concentration process. Such process would vary in intensity from lake to lake leading to different degrees of elemental concentration and of secondary salt mineral precipitations. The region investigated in the present study is indeed under hypercontinental climate (e.g., Sklyarov et al., 2010; Vogt et al., 2010). Such a climate is dry and can therefore induce a net elemental concentration process due to evaporation in summer and/or to cryogenesis in winter.

The assumption that the lake water salinity is mainly controlled by concentration processes was investigated using the EQL/EVP program (Risacher and Clément, 2001). This numerical code has been computed to simulate the evaporation of freshwaters to high concentration at temperatures equal or above 0 °C. The thermodynamical parameters controlling the dissolution and precipitation of natural minerals are indeed usually given or determined for temperatures above 0 °C. The results of such simulation can be therefore questioned as in this Siberian region mean monthly temperatures can significantly drop under 0 °C, especially in winter. However, as the mean annual temperature remains close to 0 °C, we consider that the results of the simulation performed at 0 °C can be seen as a first and realistic approximation.

The code calculates the chemical composition of waters during an evaporation process, using activity coefficients based on the Pitzer's interaction model, which makes possible their calculation to high ionic strength. The set of mass-action equations is solved through a Newton-Raphson method (for details, see Risacher and Clément, 2001). The results of an evaporation process, occurring at 0 °C at equilibrium with atmosphere, for the five spring waters collected and analyzed in this study are given in Fig. 5, where the concentrations of the major cations are plotted against the chloride concentrations. The data points representative of the lake waters are also plotted on the diagrams. The numerical simulations were run until the salinity reaches 100 g/L, the highest order of salinity measured in the study lakes. The simulations point out that intense evaporation/concentration of spring waters leads to the precipitation of first Ca and Mg-carbonates, then of sulfate minerals such as gypsum and even mirabilite for a salinity higher than 50 g/L (Fig. 5). This is in agreement with the occurrence of carbonates and gypsum reported by Sklyarov et al. (2010) in a sediment core from a saline lake in the same area, and with the observation of mirabilite in the hollow of dried-up lakes in the Barguzinskaya Rift zone (Akulov and Rubtsova, 2011), a geographical area very close to the study region, and hence submitted to a very similar climate and hydrological working. Furthermore, it appears from Fig. 5 that such a simple evaporation scenario accounts relatively well for the geochemical characteristics of the lake waters. In all the X versus Cl logarithmic diagrams, the data points of lake waters plot to the first order along the evaporation trends of the different spring waters. All these observations certainly argue for the simple ‘concentration/evaporation’ scenario we propose to account for deriving the geochemical characteristics of the lake waters from spring type waters.

Two main parameters have to be taken into account in order to explain the coexistence of lakes with low salinity and others with higher salinity in the same geographical area. The first concerns the input of feeding water into the lakes as compared with the water loss through evaporation or any other concentration process. As the climate is homogeneous at the scale of the study area, the relevant parameter in this case would be the input of feeding water into the lakes. Because the diluted lakes and the saline ones are located in slightly different structural settings, one might assume that the input of feeding waters would be more important for diluted lakes than for saline ones. The other parameter to take into account for explaining the range of salinity of lake waters observed in the Ol’khon Region, is the duration of the concentration process. If such a time parameter were the key parameter, this would mean that saline lakes have undergone a longer period of water concentration than the diluted ones. Field observations and lake sediment datings indicate that the age of the lakes can be variable at the scale of the study area. In the Ol’khon Region, it was observed, for instance, that some lakes dried out for quite long periods of time, and that, new lakes were formed during the past 10 or 20 years (Sklyarova et al., 2002; Sklyarov, pers. comm.). Furthermore, ¹⁴C-datings analyses have pointed to ages as old as 8000 years for lacustrine sediments of a saline lake of the Ol’khon Region (Sklyarov et al., 2010), indicating that some of these lakes can be old.

U isotopic data show that time is probably an important parameter to be taken into consideration for explaining the large range of lake salinities. As shown in the result section, the (²³⁴U/²³⁸U) activity ratios of the most concentrated lake waters are higher than those of the spring and diluted lakes in contrast to Sr isotope ratios which exhibit very similar ranges of variations for diluted and salt lake waters. Such a difference can be the consequence of a longer period of water-rock interactions between lake waters and lake sediments in the concentrated lakes than in the diluted ones. Indeed during water-rock interactions the behavior of ²³⁴U and ²³⁸U isotopes is not similar, i.e., ²³⁴U tends to be more enriched in waters than ²³⁸U. This is explained by the well-known alpha-recoil process: during the alpha radioactive decay of ²³⁸U into ²³⁴Th, which decays very quickly into ²³⁴U, the recoil energy can be sufficient so to eject the ²³⁴Th isotope outside the crystal lattice, or if not ejected, the ²³⁴Th isotope is in a damaged lattice compared to that initially occupied by the parent nuclide ²³⁸U. Thus, during the water-rock interactions ²³⁴U is either directly ejected into the surrounding water, or becomes more mobile than ²³⁸U (for more details, see Ivanovich and Harmon, 1992; or Chabaux et al., 2003; 2008). Consequently, for similar physical characteristics (same water-rock ratios, same grain size, etc.) the water enrichment in ²³⁴U compared to ²³⁸U will be all the more greater than the duration of water-rock interaction will be longer. Assuming, for the studied lakes that the enrichment of ²³⁴U in water is only controlled by the ejection of ²³⁴U out of the crystal lattice of the sediment minerals, then the (²³⁴U/²³⁸U) activity ratios in lake waters will obey the following relationship (Henderson et al., 1999):

In the case of sediment and water of the study lakes, the following parameters can be proposed: r = 10 μm, α = 0.055 μm, ²³⁸U_sed = 80 ppm (Sklyarov et al., 2010), ²³⁸U_water = 5 ppb i.e., the mean value in spring waters, ρ_sed = 2.8 g/cm³ (as carbonates and biotite are the main minerals of the lake sediments, Sklyarov et al., 2010), ρ_water = 1 g/cm³, φ = 0.7 (Sklyarov, pers. comm.), λ₂₃₄ = 2.83 × 10⁻⁶ y⁻¹. Using these values a time of about 4000 years is calculated for the (²³⁴U/²³⁸U) activity ratio of lake waters to increase from about 2.2 (i.e., the highest U activity ratio in spring and diluted lake waters) to 3.0 (i.e., the U activity ratio of the most concentrated lake waters). Such an age is a minimum age, as the calculation implicitly assumes that all the lake water is in contact with the sediment. This age is also very dependent on the mean values retained for the model parameters, especially the water and sediment concentrations in U, and the sediment porosity. For instance, the age will range from 2400 yr to 6400 yr when the mean sediment porosity varies from 0.6 to 0.8%. It will vary from 2100 yr to 5500 yr for a variation of the U concentration in feeding waters from 3 to 7 ppm and from 2900 yr to 4900 yr for a U concentration in the sediments ranging from 60 to 100 ppm.

This age estimate remains therefore a first order approximate. It is, however, important to notice at this stage that it is entirely consistent with the age of 8000 years inferred for the sediments of a saline lake in the same geographical area (Sklyarov et al., 2010). Such a consistency can be considered as a first argument indicating that the duration of the concentration by evaporation process can be a parameter playing a leading part in explaining the salinity range of the lake waters in the Ol’khon Region. If it were assumed that the water flux feeding the different lakes had similar geochemical characteristics, it might be then concluded that the most saline lakes in the Baikal Region would be the oldest ones. Assuming similar concentration processes of the lake waters (evaporation in summer, cryogenesis in winter) at the scale of the study area, such a relationship between the age of the lakes and salinity of their waters might indicate that the water supply of the saline lakes has been stable for a much longer time period than that of the freshwater lakes. The two types of lakes are located in slightly different tectonic settings (Section 2). Therefore one should assume, in the above scenario, that the hydrology of these two tectonic contexts evolved differently over the recent past. In turn recent tectonic movements could induce changes in soils and/or bedrocks associated or not to spatial variations of the frozen soil thicknesses. At this stage, such assumptions remain clearly working hypotheses, which need to be further discussed on the basis of further investigations focused on the hydrological and hydrogeological contexts of the region and/or on the study of salt lake sediments. This relationship between the age of the lake and their salinity would indeed disappear if the different lakes were fed by waters with different geochemical characteristics, especially for their U concentrations. Nevertheless, it is important to stress that the U data obtained on the lakes of the Ol’khon region confirm that these lakes can be quite old and hence the lake sediments relevant paleoenvironmental recorders for the last few ky. Due to the abundance of lakes in the Lake Baikal Region, the geochemical study of their bottom sediments could help constraining the response of the Lake Baikal Region to Quaternary climatic variations, at a much finer spatial scale than that obtained up to now by analyzing only sedimentary deposits of the Baikal Lake. Such an information is relevant to fit, at a regional scale, the results of global climatic models and to generate a much more precise regional modeling of the response of the past, present and future climate changes in the Baikal Lake Region.