Hydrogeochemical Study on Closed-Basin Lakes in Cold and Semi-Arid Climates of the Valley of the Gobi Lakes, Mongolia: Implications for Hydrology and Water Chemistry of Paleolakes on Mars

: Previous studies suggested that, generally, the climate of early Mars would have been semi-arid when the surface temperatures were above freezing. On early Mars, closed-basin lakes would have been created; however, the hydrogeochemical cycles of the lake systems are poorly constrained. Here we report results of our ﬁeld surveys to terrestrial analogs of closed-basin lake systems that developed in cold and semi-arid climates: The Valley of the Gobi Lakes of Mongolia. Our results show that groundwater plays a central role not only in hydrology, but also in geochemical cycles in the lake systems. We ﬁnd that groundwater predominantly ﬂows into the lakes through local seepage and regional ﬂows in semi-arid climates. Through the interactions with calcite-containing soils, local groundwater seepage provides Ca 2 + and HCO 3 − to the lakes. In the wetland located in between the lakes, high-salinity shallow pools would provide Cl − and Na + to the groundwater through inﬁltration. If similar processes occurred on early Mars, local seepage of groundwater would have provided magnesium and alkalinity to the early Jezero lakes, possibly leading to authigenic precipitation of lacustrine carbonates. On early Mars, inﬁltration of surface brine may have transported salts and oxidants on the surface to lakes via regional groundwater ﬂows. We suggest that inﬂows of multiple types of groundwater in semi-arid climates could have caused redox disequilibria in closed-basin lakes on early Mars.

. A geographical map of Mongolia and the locations of the Valley of the Gobi Lakes (the red ellipse with dotted line), based on [28]. The research area in the present study is outlined by the black square. A satellite image of this targeted area is shown in Figure 2 below. Major rivers (the bold curves) and lakes (the black areas) are also shown. Mountain ranges (the Khangai and Altai mountains) are also represented as shadows. Based on [32], a simplified map of Kӧppen-Geiger climate classification zones is superposed on the geographical map.

Geohydrological Setting
The study area is the Valley of the Gobi Lakes in central Mongolia ( Figure 1). The Valley of the Gobi Lakes exists in a latitudinally oriented depression of tectonic origin, lying in between the Khangai and Altai Gobi mountains ( Figure 1). Two major lakes-the Böön Tsagaan lake and the Orog lake-are located in the middle of the Valley of the Gobi Lakes ( Figure 1). The elevations from sea level of the Böön Tsagaan and Orog lakes are ~1305 m and ~1220 m, respectively [25]. The Böön Tsagaan and Orog lakes are fed by the inlet rivers of the Baidrag and Tuin rivers, respectively, whereas there are currently no outlet rivers from the lakes (Figure 2). Both of the Baidrag and Tuin rivers originate in the Khangai mountains ( Figure 1). The Khangai mountains reach above 3000 m in elevation and contain continuous and discontinuous permafrost [26]. These rivers continue through proluvial regions of dry steppe and end in the Böön Tsagaan and Orog lakes in the semi-arid climate zone. The present-day lake surface areas are ~250 km 2 for Böön Tsagaan lake and ~140 km 2 for Orog lake, with the maximum depths of 10 m and 3 m, respectively [26]. The catchment areas of the Böön Tsagaan-Baidrag and Orog-Tuin systems are ~45,000 km 2 and ~14,900 km 2 , respectively [26]. In the catchment areas near the Khangai mountains, there are a number of small closed-basin lakes, including the Olgoy lake, at topographic lows in association with melting of sporadic ground ice and snow ( Figure 2). The Olgoy lake is likely an oxbow lake of the nearby Sumber river, which is a tributary of the Baidrag river. There are currently no inlet and outlet rivers of Olgoy lake. There are also small depressions within the valley, which may have been lakebeds in the past [28]. These small depressions include the Tsagaan Nuur wetland (Figure 2), where lacustrine sediments and shallow pools are present. The Tsagaan Nuur wetland belongs to the catchment area of the Böön Tsagaan lake [26].  [28]. The research area in the present study is outlined by the black square. A satellite image of this targeted area is shown in Figure 2 below. Major rivers (the bold curves) and lakes (the black areas) are also shown. Mountain ranges (the Khangai and Altai mountains) are also represented as shadows. Based on [32], a simplified map of Köppen-Geiger climate classification zones is superposed on the geographical map.
In the field survey we measured water chemistry of the lakes, rivers, springs, and wells to understand the hydrogeochemical cycles that support the lacustrine systems. Based on the results, we discuss the implications for the hydrogeochemical cycles of closed-basin lake systems that developed in semi-arid climates on early Mars.

Geohydrological Setting
The study area is the Valley of the Gobi Lakes in central Mongolia ( Figure 1). The Valley of the Gobi Lakes exists in a latitudinally oriented depression of tectonic origin, lying in between the Khangai and Altai Gobi mountains ( Figure 1). Two major lakes-the Böön Tsagaan lake and the Orog lake-are located in the middle of the Valley of the Gobi Lakes ( Figure 1). The elevations from sea level of the Böön Tsagaan and Orog lakes are~1305 m and~1220 m, respectively [25]. The Böön Tsagaan and Orog lakes are fed by the inlet rivers of the Baidrag and Tuin rivers, respectively, whereas there are currently no outlet rivers from the lakes (Figure 2). Both of the Baidrag and Tuin rivers originate in the Khangai mountains ( Figure 1). The Khangai mountains reach above 3000 m in elevation and contain continuous and discontinuous permafrost [26]. These rivers continue through proluvial regions of dry steppe and end in the Böön Tsagaan and Orog lakes in the semi-arid climate zone. The present-day lake surface areas are~250 km 2 for Böön Tsagaan lake and~140 km 2 for Orog lake, with the maximum depths of 10 m and 3 m, respectively [26]. The catchment areas of the Böön Tsagaan-Baidrag and Orog-Tuin systems are~45,000 km 2 and~14,900 km 2 , respectively [26]. In the catchment areas near the Khangai mountains, there are a number of small closed-basin lakes, including the Olgoy lake, at topographic lows in association with melting of sporadic ground ice and snow ( Figure 2). The Olgoy lake is likely an oxbow lake of the nearby Sumber river, which is a tributary of the Baidrag river. There are currently Minerals 2020, 10, 792 4 of 26 no inlet and outlet rivers of Olgoy lake. There are also small depressions within the valley, which may have been lakebeds in the past [28]. These small depressions include the Tsagaan Nuur wetland (Figure 2), where lacustrine sediments and shallow pools are present. The Tsagaan Nuur wetland belongs to the catchment area of the Böön Tsagaan lake [26].
Minerals 2020, 10, x FOR PEER REVIEW 4 of 28 The Khangai mountains consist mainly of Precambrian and Paleozoic sedimentary rocks and Mesozoic granitic intrusions, whereas the Altai Gobi mountains is formed by transpressional Cainozoic movements of Paleozoic volcanic and sedimentary rocks [33]. Major faults exist along with the Khangai and Altai Gobi mountains, intersecting the Baidrag and Tuin rivers [33]. The Valley of the Gobi Lakes is an elongated inter-montane depression between the Siberian craton and the Tarim craton [33]. The basins of the Valley of the Gobi Lakes are filled with Quaterary alluvial fans, lacustrine sediments, and aeolian deposits [28,33]. The alluvial, lacustrine, and aeolian deposits are poorly cemented. The thickness of the alluvial, lacustrine, and aeolian deposits can reach 20-40 m or greater in the Valley of the Gobi Lakes, capable of storing groundwater in the subsurface.  (Table 1), being a topographic barrier. (c) The locations of the sampling points around the Olgoy lake. The colors of the sampling locations of groundwater represent the types of water chemistry (see the annotation in the figure and Section 4.1 below). More detailed information on the locations are summarized in Table 1. (d) A simplified topographic map of our research area, corresponding to the area shown in the panel (a). The original data were derived from GSI Maps of Geospatial Information Authority of Japan.
The mean annual precipitation of the Valley of the Gobi Lakes is 50-100 mm. In the last thirty years, the mean annual air temperature in this area has increased by 1-2 C upon global warming,  (Table 1), being a topographic barrier. (c) The locations of the sampling points around the Olgoy lake. The colors of the sampling locations of groundwater represent the types of water chemistry (see the annotation in the figure and Section 4.1 below). More detailed information on the locations are summarized in Table 1. (d) A simplified topographic map of our research area, corresponding to the area shown in the panel (a). The original data were derived from GSI Maps of Geospatial Information Authority of Japan.
The Khangai mountains consist mainly of Precambrian and Paleozoic sedimentary rocks and Mesozoic granitic intrusions, whereas the Altai Gobi mountains is formed by transpressional Cainozoic movements of Paleozoic volcanic and sedimentary rocks [33]. Major faults exist along with the Khangai and Altai Gobi mountains, intersecting the Baidrag and Tuin rivers [33]. The Valley of the Gobi Lakes is an elongated inter-montane depression between the Siberian craton and the Tarim craton [33]. The basins of the Valley of the Gobi Lakes are filled with Quaterary alluvial fans, lacustrine sediments, and aeolian deposits [28,33]. The alluvial, lacustrine, and aeolian deposits are poorly cemented. The thickness of the alluvial, lacustrine, and aeolian deposits can reach 20-40 m or greater in the Valley of the Gobi Lakes, capable of storing groundwater in the subsurface.
The mean annual precipitation of the Valley of the Gobi Lakes is 50-100 mm. In the last thirty years, the mean annual air temperature in this area has increased by 1-2 • C upon global warming, enhancing melting of permafrost in the mountain areas [28]. Although the annual mean temperature of this area is around 1 • C, there are significant regional differences. For instance, the annual mean temperature around the Olgoy lake is 4-5 • C lower than that of the Orog and Böön Tsagaan lakes. Precipitation usually occurs predominantly from July to September with a mean precipitation of 10-50 mm/month, whereas the other seasons are cold and arid with precipitation less than 5 mm/month. The aridity index (the ratio of precipitation to potential evaporation) is estimated to be 0.2-0.3 in the Valley of the Gobi Lakes [25]. The precipitation in the Valley of the Gobi Lakes is insufficient to support the water volumes of both of the Böön Tsagaan and Orog lakes [26], suggesting the importance of melting snow and permafrost in the Khangai mountains supporting the lake systems with a water supply [27]. The mean precipitation around the Olgoy lake area near the mountains (~200 mm/year) is higher than that of the Orog and Böön Tsagaan lakes. The surface area of the Böön Tsagaan lake has been on monotonic decrease over the last twenty years based on remote sensing data, whereas the Orog lake has dried up a few times in the last 40 years [26]. This may have happened due to a combination of several factors, including changes in precipitation and a recent increase in evaporation [26].

Geological Survey and Sampling
We performed geological surveys for the Valley of the Gobi Lakes and its catchment areas in October 2017 (the beginning of a dry season) and August 2018 (the middle of a rainy season). Water samples were collected from the Böön Tsagaan, Orog, and Olgoy lakes as well as surrounding wells and springs. Sediment samples were also collected from the lake floors and the surrounding areas of the Valley of the Gobi Lakes (see below).
Water samples: Water samples were taken from the Böön Tsagaan, Orog, and Olgoy lakes and the Baidrag, Tuin, and Sumber rivers, 19 wells, and 2 springs surrounding the lakes ( Figure 2 and Table 1). Water samples from wells and springs are generally believed to represent local groundwater chemistry. In the 2018 rainy season survey, we collected a water sample from a shallow pool in the Tsagaan Nuur wetland. Dissolved O 2 (DO) levels, electric conductivity (EC), oxidation-reduction potential (ORP), pH, alkalinity, and temperature of water samples were measured on site (see below for the measurements). For the chemical analysis of dissolved species, the collected water samples were stored in two pre-cleaned plastic bottles with volumes of 100 mL after filtering with a 0.45-µm membrane (see below for the chemical analysis). We also measured the water levels at wells at the timing of sampling in 2017 (Table 1).
Sediment samples: We collected sediment samples in the area around the Böön Tsagaan lake and Tsagaan Nuur wetland ( Figure 2). Four samples (OC 1-4) were collected from outcrops around the Böön Tsagaan lake ( Figure 2). Present-day lacustrine sediment samples were collected from the Böön Tsagaan lake in 2016, from the Orog lake in 2015, and from the Olgoy lake in 2019 in our previous surveys. The pictures of the outcrops are summarized in Supplementary Figure S1.
The samples of OC 1-3 were collected from the locations near proposed paleo-shorelines of larger lakes in more humid climates in the past [28]. The altitudes of OC 1-3 were close to each other (~1280 to 1290 m above sea level), suggesting that they were deposited at almost the same horizon. At the outcrops, poorly cemented laminated sandy sediments were commonly observed (Supplementary Figure S1), supporting the hypothesis that their formation was due to past water activities. Given the presence of conglomerate layers within the sandy sediments (Supplementary Figure S1), these sediments may be flood plain deposits formed by repeated floods in the past. The sample of OC 4 was collected from a higher altitude location (1325 m for OC 4 above sea level). The presence of laminated sandy sediments in the outcrop of OC 4 suggests that the water levels of larger paleolakes may have reached a minimum of this altitude (~1325 m) in the past. In the outcrop of OC 4, we sometimes find carbonate concretions 0.1-1 cm in diameter in the subsurface soils (Supplementary Figure S1). We collected~10 to 50 g of sediment samples from these outcrops.
The lacustrine sediment samples within the Böön Tsagaan and Orog lakes were taken using a gravity core sampler. The gravity core sampler-with an inner tube 1 m in length and 60 mm in diameter-was thrown into the lakes from a boat and it then penetrated the unsolidified sediments. The lacustrine sediment samples used in the present study were located at a depth 10 cm from the lake floor. During winter, the surface of Olgoy lake is covered with thick ice. We collected the lacustrine sediment sample from the bottom of the frozen lake by drilling through the surface ice in February 2019. All of the lacustrine sediments were freeze-dried in a laboratory after collection.

Chemical and Mineralogical Analyses
Water chemistry: For the on-site DO measurements of the water samples, we employed a DO meter (DO-3IP; TOA-DKK Corp. Milton, Cambridge, UK). We used a portable electrical pH meter with electrodes (WM-32EP; TOA-DKK Corp.) for the EC, pH, ORP, and temperature measurements on site. Prior to the measurements, the pH electrode was calibrated with pH buffer solutions at pH 4.01, 6.86, and 9.18. After filtering using a 0.45-µm membrane, total alkalinity was also measured on site for the water samples through titration using sulfuric acid (H 2 SO 4 ) solutions.
We collected two bottles of the water samples at each sampling site. One was used for the measurements of dissolved total Na, K, Ca, and Mg. Ultra-pure nitric acid (HNO 3 ) was added to this bottle so that the HNO 3 concentration in the water sample became 0.6%. The water sample in the other bottle was left for the measurements of dissolved Cl − and SO 4 2− . These major dissolved species in the water samples were measured using an inductively coupled plasma optical emission spectrometer (ICP-OES) (ES-710; Varian Inc. Palo Alto, CA, USA). To calculate the species concentrations (speciation) and the saturation index of some secondary minerals (e.g., calcite and gypsum) in the water samples, a chemical equilibrium calculation code of the REACT program in Geochemist's Workbench [34] was employed. In the present study, we evaluated whether evaporation of the inlet river water and groundwaters can reproduce the compositions of lake waters (see Section 5.1 below). To this end, the REACT program in Geochemist's Workbench was also run to calculate the evaporative concentrations of dissolved species and consequent precipitation of secondary minerals by progressively removing pure water abundance from the measured river water and groundwaters at temperature of 20 • C. In the calculations of evaporative concentrations, we did not take into account any removal of Na + and K + . In reality, these cations could be removed from lake water through adsorption onto clay minerals.
Mineralogical composition and grain size distribution: To obtain the mineralogical compositions of the sediment samples, X-ray diffraction (XRD) spectra were obtained for the dried sediment samples with an X-ray diffractometer (X'PERT-PRO; PANalytical, Malvern, UK). Prior to the XRD analysis, 0.5 g of the sediment samples was pounded in an agate mortar. The scan range is 3.011 • -64.993 • in 2θ. The generator voltage is 45 V. The scan step is 0.0167 • . In addition to the XRD analysis, we performed grain-size analysis for the sediment samples using a He-Ne laser-diffraction-scattering grain size analyzer (SALD2200J; SHIMADZU Corp. Kyoto, Japan). Approximately 0.05 g of the sediment samples were first dispersed with an ultrasonic wave in 50 mL of sodium hexametaphosphate solutions for 5 min. The solutions with the sediment samples were then emptied into the analyzer with~20 mL of sodium hexametaphosphate solution. The concentrations of the mixed solutions were adjusted to obtain a sufficient laser transmission level around 40-60%. The sediments in the solutions were dispersed with an ultra-sonic wave in a transport circuit accompanying the analyzer. The grain-size distributions were obtained for a diameter range between 0.03 and 1000 µm. Tables 2 and 3 summarize pH, DO, ORP, alkalinity, EC, and the compositions of major elements of collected water from lakes, rivers, springs, and wells. The hexa-diagrams and piper diagrams of water chemistry in meq/kg are also shown in Figures 3 and 4, respectively. The lake waters exhibit different water chemistry to the rivers, springs, and wells (Tables 2 and 3). The lake waters are all alkaline (pH 9-9.5) and hyposaline (salinity 30-200 mmol/kg); meanwhile, those of the rivers, springs, and wells are circumneutral (pH 6.4-8.1) and low salinity (Tables 2 and 3). The lake waters are categorized as a Na-Cl type for Böön Tsagaan lake, which means that the major cation and anion are Na + and Cl − , respectively (Tables 2 and 3, and Figure 3). The waters of the Olgoy and Orog lakes are categorized as a Na-HCO 3 type (Tables 2 and 3, and Figure 3), although the Cl − and Mg 2+ concentrations are significant for the Orog and Olgoy lakes, respectively. The high Na + (and Cl − ) concentrations and alkaline pH are most likely a result of evaporation of the lake waters, leading to accumulation of conservative ions of Na + (and Cl − ) within the lakes. The high HCO 3 − concentrations in the three lakes are achieved by effective dissolution of atmospheric CO 2 into alkaline lake water. Compared with the rivers, springs, and wells, the relative concentrations of Ca 2+ are low in the lakes because it is removed as precipitates of carbonate minerals (calcium carbonates) within the lakes (see below).    Tables 2 and 3 were converted to meq/kg.  Tables 2 and 3 were converted to meq/kg.

Water Chemistry of Lakes, Rivers, and Groundwater
A review of the water chemistries of 24 alkaline lakes with pH > 9 on Earth shows that the water chemistries are apparently equilibrated with respect to monohydrocalcite (MHC) (CaCO 3 ·H 2 O) and amorphous Mg carbonate (AMC) (MgCO 3 ·nH 2 O) [35]. The saturation indexes of MHC of the Böön Tsagaan, Orog, and Olgoy lakes are commonly slightly supersaturated with respect to MHC from 0.3 to 0.6, suggestive of precipitation of MHC in the lakes. The precipitations of MHC were confirmed through collection and analysis of suspended materials in these lakes by our companion paper [36]. The saturation indexes of AMC of the lakes during a rainy season in 2018 range from -0.1 to −0.6, although those during a dry season in 2017 are slightly low-down to −1.0-compared with those in the wet season (the seasonal difference in the saturation indexes cannot be explained by the difference in surface temperatures). Nevertheless, the saturation indexes of AMC in the lakes are significantly higher than those of the rivers, springs, and wells. The previous study on laboratory experiments and geochemical modeling indicated that the co-existence of AMC in solutions is required for precipitation of MHC [37]. Thus, the results of the near-zero saturation indexes of AMC (−0.1 to −0.6) suggest that Ca 2+ and Mg 2+ in the lakes may be removed by formation of AMC, in addition to MHC, during the evaporation of lake water.
In contrast to the lake waters, we find that the rivers, springs, and wells generally have low salinity (Tables 2 and 3). Our results of the major dissolved components indicate that the water chemistry of groundwaters (wells and springs) can be classified into a few groups (Figures 3 and 4).  Table 1). In contrast, the hexa−diagrams show that the groundwaters close to Tsagaan Nuur wetland contain Na + , Mg 2+ , Cl − , and SO 4 2− in addition to Ca 2+ and HCO 3 − (Figure 3d) (hereafter, we call them group B). The groundwater of group B is distributed east of the group A wells, toward Tsagaan Nuur wetland (Figure 2b). The groundwater of the west shore of the Böön Tsagaan lake (the well of BTS well 1 west) can also be categorized as group B (Figure 3d). Although the number of available sampling sites is limited, the groundwater of the west shore of the lake may contain more SO 4 2− than the other groundwaters of group B. The Baidrag river may be categorized as an intermediate between groups A and B, which means that the water chemistry of the Baidrag river is a Ca−HCO 3 type, the same as group A, but contains a significant fraction of SO 4 2− and Mg 2+ (Figure 3b). The elevation of the water tables of group−B groundwater (1280-1305 m) are generally lower than those of group−A groundwater (Table 1).
In the piper diagram, the groundwater of group B can be divided into two sub−groups ( Figure 4): in one, the groundwater contains relatively high SO 4 2− and low Cl − (e.g., the wells of BTS−TN 20: Figure 4a), and in the other, the groundwater contains relatively high Na + and/or Cl − (e.g., the wells of TN 25 and TN 29: Figure 4). We believe that the groundwaters of TN 22 and TN 27 can be categorized as the latter subgroup given the high Na in the groundwaters (Figure 4a). The groundwaters of the former subgroup seem to be distributed not only west of Tsagaan Nuur wetland (i.e., BTS−TN 20), but also near the Böön Tsagaan lake (i.e., BTS spring and BTS west) (hereafter, we call this subgroup group B−west) ( Figure 2b); whereas the wells of the latter subgroup are located near Tsagaan Nuur wetland (hereafter, we call them group B−east) (Figure 2b). The elevation of the water table of group−B groundwater tends to be lower toward the east and becomes lowest near Tsagaan Nuur wetland (~1280 m at TN 27 in Figure 2b) Tables 2 and 3 were converted to meq/kg.
Interestingly, the groundwater close to the Bӧӧn Tsagaan, Olgoy, and Orog lakes has water chemistry that is distinct from the lake waters (Figures 2 and 3). The distances from the wells to each adjacent lake are a few km or less (Figure 2). The water tables at these wells are a few meters below the surface, which are comparable to or higher than the respective lake levels (Table 1). If the lake water outflows toward the surrounding subsurface, the groundwater adjacent to the lakes should be similar to that of the lakes; namely, a Na−Cl type with high salinity. Thus, the low salinity and  Tables 2 and 3 were converted to meq/kg. Similar variations in groundwater chemistry can be found in the areas around both of the Olgoy and Orog lakes. The water chemistry of the Sumber river and some groundwaters around the Olgoy lake is Ca−HCO 3 type (see Figure 3b,c and Figure 4c,d), similar to group A of the Böön Tsagaan lake (Figure 3). In contrast, the groundwaters east and south of the Olgoy lake contain Na + , Cl − , Mg 2+ , and SO 4 2− in addition to Ca 2+ and HCO 3 − (see Figure 3c and 4), similar to group B−west ( Figure 3).
The groundwater and Tuin river of the Orog lake are Ca−HCO 3 type (group A) ( Figure 3). As far as we know from our investigation, there is no groundwater similar to group B−east around the Orog and Olgoy lakes. The measured variety of groundwater chemistry suggests that the groundwater compositions are largely determined by local hydrogeochemical processes, rather than the composition of adjacent lakes. The causative mechanisms of this various water chemistry are discussed below in Section 5.2. Interestingly, the groundwater close to the Böön Tsagaan, Olgoy, and Orog lakes has water chemistry that is distinct from the lake waters (Figures 2 and 3). The distances from the wells to each adjacent lake are a few km or less (Figure 2). The water tables at these wells are a few meters below the surface, which are comparable to or higher than the respective lake levels ( Table 1). If the lake water outflows toward the surrounding subsurface, the groundwater adjacent to the lakes should be similar to that of the lakes; namely, a Na−Cl type with high salinity. Thus, the low salinity and Ca−HCO 3 type of the groundwaters near the lakes strongly suggest that there is no effective outflowing of the lake water, despite saline water having high density. There are two possible explanations for these results: In one, the interactions between the lake and surrounding groundwater are highly inhibited due to fine−grain impermeable lake sediments (Section 4.2), isolating the lake water from the surrounding groundwater. In the other, the surrounding groundwater is predominantly flowing into the lakes without any significant outflowing of the lake water. These are also discussed below in Section 5.1.
In addition to spatial variations in water chemistry, we find a small temporal variation in the water chemistry of some of the rivers, wells, and springs, comparing the results from the survey during a dry season of 2017 and a rainy season of 2018 ( Figure 3). For instance, the Baidrag river water and the groundwater of group A around the Böön Tsagaan lake tend to contain less Ca 2+ and HCO 3 − during a dry season (fall in 2018) than during a rainy season (summer in 2017). The groundwater chemistry of some of group A becomes close to group B−west in a dry season (Figure 4). A similar trend can be seen for some of the springs and groundwaters of the Olgoy and Orog lakes (Figure 3). This is probably because dissolution of carbonate minerals (e.g., calcite) in the soils was enhanced during the rainy season, which results in the relatively high concentrations of Ca 2+ and HCO 3 − in the groundwaters and rivers. However, the seasonal changes do not affect the water chemistry remarkably compared with the observed spatial variations of the groundwater chemistry ( Figure 4). Figure 5 shows the results of our XRD analysis for the sediment samples. The samples from the outcrops (OC 1-4) are predominantly composed of anhydrous silicates of albite and quartz ( Figure 5). Low levels of a broad peak at~20 to 30 • indicate that there are low abundances of amorphous phases and organic matter in the samples ( Figure 5). These results indicate that the outcrop samples, in general, experience only low degrees of aqueous alterations, although sample OC 4 may contain a small amount of hydrated phyllosilicate of dioctahedral smectite (e.g., montmorillonite) ( Figure 5). Although we found calcite concretion in the soils (Supplementary Figure S1), no peaks of carbonate minerals are found in the XRD spectra of the outcrop samples, suggesting the calcite contents would be too low to be detected.

Minerals and Grain Size of Lake Sediments
In contrast to the outcrop samples, the lake sediment samples of the Böön Tsagaan, Olgoy, and Orog lakes contain abundances of secondary minerals, such as calcite and MHC, in addition to primary albite and quartz ( Figure 5). Relatively high levels of a broad peak at~20 • to 30 • show high abundances of amorphous phases and/or organic matter in the lake sediment samples ( Figure 5). Illite and chlorite found in the lake sediments likely have detrital origins given their narrow XRD peaks, and thus high crystallinity. Illite and chlorite may be detrital components of hydrothermally altered metamorphic basement rocks in the catchment area. In contrast to the outcrop samples, the lake sediment samples of the Bӧӧn Tsagaan, Olgoy, and Orog lakes contain abundances of secondary minerals, such as calcite and MHC, in addition to primary albite and quartz ( Figure 5). Relatively high levels of a broad peak at ~20° to 30° show high abundances of amorphous phases and/or organic matter in the lake sediment samples ( Figure 5). Illite and chlorite found in the lake sediments likely have detrital origins given their narrow XRD peaks,  Table S1 summarized the major minerals and their XRD intensities. Figure 6 shows the results of our grain−size analysis for the outcrop and lake sediment samples. The obtained grain−size distributions of the outcrop samples (OC 1-4) are unimodal at a mean diameter of 100-300 µm, except for a faint peak at 0.3 µm for OC 2. In contrast, the lake sediment samples exhibit a multi−modal distribution of grain size at one peak diameter of 5 µm and the other of 30-50 µm. Given the unimodal grain size distributions of the outcrop samples at 100-300 µm, the peak with a mean diameter of 30-50 µm of the lake sediments may be the detrital component of non−hydrous albite and quartz. On the other hand, the grains with mean diameter of 5 µm would be due to the secondary minerals, such as carbonate minerals, found in the XRD results. metamorphic basement rocks in the catchment area. Figure 6 shows the results of our grain−size analysis for the outcrop and lake sediment samples. The obtained grain−size distributions of the outcrop samples (OC 1-4) are unimodal at a mean diameter of 100-300 μm, except for a faint peak at 0.3 μm for OC 2. In contrast, the lake sediment samples exhibit a multi−modal distribution of grain size at one peak diameter of 5 μm and the other of 30-50 μm. Given the unimodal grain size distributions of the outcrop samples at 100-300 μm, the peak with a mean diameter of 30-50 μm of the lake sediments may be the detrital component of non−hydrous albite and quartz. On the other hand, the grains with mean diameter of 5 μm would be due to the secondary minerals, such as carbonate minerals, found in the XRD results.

Hydrogeochemical Cycles of the Hyposaline Lakes in the Valley of the Gobi Lakes
Although we performed regional−scale field surveys for the Valley of the Gobi Lakes, the sampling sites are still sparce compared to the whole of the catchment areas of the Bӧӧn Tsagaan and Orog lakes. In particular, there is only one sampling site in the west of the Bӧӧn Tsagaan lake (BTS west) and a few sampling sites around the Orog lake ( Figure 2). The water chemistry of BTS west exhibits a distinct chemical composition, which is characterized as the highest fraction of SO4 2− in the anions among the sampling sites ( Figure 4). Thus, the groundwater seepage from the west of Bӧӧn Tsagaan lake, if occurred, could provide additional SO4 2− to the lake, although SO4 2− −rich groundwater cannot be the major source of water to the lake to explain the chemistry of Bӧӧn Tsagaan lake (see Section 5.1.1 below). In addition, we performed field surveys in a rainy and dry season in

Hydrogeochemical Cycles of the Hyposaline Lakes in the Valley of the Gobi Lakes
Although we performed regional−scale field surveys for the Valley of the Gobi Lakes, the sampling sites are still sparce compared to the whole of the catchment areas of the Böön Tsagaan and Orog lakes. In particular, there is only one sampling site in the west of the Böön Tsagaan lake (BTS west) and a few sampling sites around the Orog lake ( Figure 2). The water chemistry of BTS west exhibits a distinct chemical composition, which is characterized as the highest fraction of SO 4 2− in the anions among the sampling sites ( Figure 4). Thus, the groundwater seepage from the west of Böön Tsagaan lake, if occurred, could provide additional SO 4 2− to the lake, although SO 4 2− −rich groundwater cannot be the major source of water to the lake to explain the chemistry of Böön Tsagaan lake (see Section 5.1.1 below). In addition, we performed field surveys in a rainy and dry season in different years (i.e., dry season in 2017 and rainy season in 2018). Thus, the seasonal variations in the water chemistry considered in the present study might partly include the annual variations of the water chemistry due to a change in precipitation.
To reduce the uncertainties caused by the limitation of the number of sampling sites and timing of survey, further investigations over the Valley of the Gobi Lakes in a single year would be required. Despite the uncertainties, we suggest that the following two conclusions of our field survey are robust. First is that the water chemistry of the lakes is distinct from those of the rivers and groundwaters. Second is that we find three types of water chemistry (i.e., groups A, B−east, and B−west) for the groundwaters and rivers despite the surveys in the different years. In this section, we discuss the interactions between the lake and groundwater (Section 5.1.1) and possible mechanisms to generate different types of groundwaters (Section 5.1.2).

Interactions between the Lake and Groundwater
In this subsection, we examine the interactions between the lake and groundwater using the measured variety of water chemistry as tracers of the water source. We consider, as an example, the Böön Tsagaan lake system because of the abundance of the measured data points around the lake. The water chemistry of Böön Tsagaan lake should be determined by the concentration of dissolved species of the water sources (river, springs, and wells) driven by evaporation of the lake water. There are two possibilities to explain the lake water chemistry by evaporative concentrations.
(1) If interactions between the lake and groundwaters are inhibited due to low permeability of the lake sediments, the lake's composition is determined solely by evaporative concentration of the river water.
(2) If groundwater is predominantly supplied into the lake without outflowing of the lake water, the lake's composition can be explained by evaporative concentrations of combined inlet river water and groundwater.
Using the REACT program of Geochemist's Workbench (see Section 3.2 for calculation methods), we simulated the evolutions in chemical composition of the evaporating river water and groundwater. The principal criterion for evaluation of the evaporative concentration is whether or not a simulation reproduces the measured composition of the Böön Tsagaan lake. Upon evaporation, we consider the precipitations of both MHC and AMC at the fixed ambient CO 2 pressure of 4 × 10 −4 atm. This is the case because the water chemistry (Mg 2+ and Ca 2+ concentrations) of the lakes is thought to be governed by the formation of and dissolution equilibria with MHC and AMC [35,36], and because calcite is found in the sediments of the Böön Tsagaan lake (Figure 4 and Section 3.2). Precipitation of MHC is the precursor of authigenic deposition of calcite in alkaline lacustrine environments [35]. Furthermore, our companion paper found MHC present in the suspended materials in the lake waters of the Böön Tsagaan lake [36]. Since MHC and AMC are considered to be the initial phases of precipitation of Mg− and Ca−bearing carbonate minerals [35,36], formations of magnesite and calcite are suppressed in our calculations. On the other hand, we considered Na + , K + , Cl − , and SO 4 2− were conservative (i.e., no removals from lake water via precipitation of minerals) upon evaporative concentration, which meant that no precipitations of chloride and sulfate minerals were taken into account. This is due to the concentrations of these elements being far below the saturation of chloride and sulfate minerals, including gypsum, epsomite, and mirabilite, in the lake (Tables 2 and 3). Figure 7 shows the evolutional tracks of water chemistry in evaporative concentration of the river water and groundwaters, superposed on the measured data in the piper diagrams. Our results of the calculations show that evaporation of the Baidrag river water cannot be the sole explanation for the compositions of the Böön Tsagaan lake (Figure 7). The water chemistry of the Baidrag river is plotted in the top−left quadrant of the piper diagram (Figure 7), where m Na < (m Cl + 2m SO42 ) (m X represents the activity of the components X in solution). In this quadrant, the difference of (m Cl + 2m SO42 ) to m Na (i.e., (m Cl + 2m SO42 ) − m Na ) increases upon evaporation of the river water, which in turn maintains high Mg 2+ concentrations in the evaporating water according to the charge balance. While Mg 2+ concentrations remain relatively high, the concentrations of HCO 3 − and CO 3 2− become very low through the precipitations of MHC and AMC based on the dissolution equilibria. As a result, the concentrations of Mg 2+ become high in the evaporating river water compared with the lake water, despite the Na + and Cl − concentrations becoming comparable (Figure 7). The results show that evaporative concentration of the river water cannot explain the lake water, suggesting that the contribution of groundwater is crucial for the Böön Tsagaan lake system. through the precipitations of MHC and AMC based on the dissolution equilibria. As a result, the concentrations of Mg 2+ become high in the evaporating river water compared with the lake water, despite the Na + and Cl − concentrations becoming comparable (Figure 7). The results show that evaporative concentration of the river water cannot explain the lake water, suggesting that the contribution of groundwater is crucial for the Bӧӧn Tsagaan lake system.  Supplementary Table S2. To achieve the water chemistry of the Bӧӧn Tsagaan Lake, we find that the input of group−A groundwater is important. A typical result of the water chemistry evolution track of evaporating group−A groundwater is also shown in Figure 7. Unlike the river water, the initial group−A groundwater is, in general, plotted in the lower−left quadrant in the piper diagram, where mNa > (mCl + 2mSO42) (Figure 7). In this quadrant, the difference of mNa to (mCl + 2mSO42) (i.e., mNa − (mCl + 2mSO42)) increases upon evaporation. According to the charge balance, this leads to low concentrations of Mg 2+ and Ca 2+ in the evaporating water (Figure 7). The concentration of Ca 2+ first decreases due to precipitation of MHC (Figure 7). Subsequently, the concentration of Mg 2+ dramatically decreases after precipitation of AMC (Figure 7). In contrast, the concentrations of HCO3 − and CO3 2− remain relatively high owing to both equilibrium with 400 ppm of atmospheric CO2 in alkaline solution and following the dissolution equilibria of MHC and AMC. As a result, the evaporating group−A groundwater tends to possess low Mg 2+ and moderate HCO3 − , which is, in general, consistent with the characteristics of the lake water (Figure 7).
One notable discrepancy of evaporating group−A groundwater is the low levels of Cl − , or low Cl − /SO4 2− , compared with the Bӧӧn Tsagaan lake water chemistry (Figure 7). Additions of the Baidrag  Supplementary Table S2. To achieve the water chemistry of the Böön Tsagaan Lake, we find that the input of group−A groundwater is important. A typical result of the water chemistry evolution track of evaporating group−A groundwater is also shown in Figure 7. Unlike the river water, the initial group−A groundwater is, in general, plotted in the lower−left quadrant in the piper diagram, where m Na > (m Cl + 2m SO42 ) (Figure 7). In this quadrant, the difference of m Na to (m Cl + 2m SO42 ) (i.e., m Na − (m Cl + 2m SO42 )) increases upon evaporation. According to the charge balance, this leads to low concentrations of Mg 2+ and Ca 2+ in the evaporating water (Figure 7). The concentration of Ca 2+ first decreases due to precipitation of MHC (Figure 7). Subsequently, the concentration of Mg 2+ dramatically decreases after precipitation of AMC (Figure 7). In contrast, the concentrations of HCO 3 − and CO 3 2− remain relatively high owing to both equilibrium with 400 ppm of atmospheric CO 2 in alkaline solution and following the dissolution equilibria of MHC and AMC. As a result, the evaporating group−A groundwater tends to possess low Mg 2+ and moderate HCO 3 − , which is, in general, consistent with the characteristics of the lake water (Figure 7). One notable discrepancy of evaporating group−A groundwater is the low levels of Cl − , or low Cl − /SO 4 2− , compared with the Böön Tsagaan lake water chemistry (Figure 7). Additions of the Baidrag river cannot increase Cl − /SO 4 2− because they are also characterized as low Cl − /SO 4 2− (Figure 7). Given the limited data of the groundwater from the west shore of the lake, we cannot conclude whether the groundwater in the west shore of the Böön Tsagaan lake can be a source of Cl−containing water or not. However, to our best knowledge, our results suggest that the groundwater in the west shore of the lake (i.e., BTS west) cannot be a source of Cl−containing water because they are also characterized as low Cl − /SO 4 2− (Figure 3). In addition, to form Cl−containing water, we suggest that infiltration of Na−Cl−rich surface water in shallow pools would be important (see Section 5.1.2), yet there are no widespread wetlands in the west shore of the Böön Tsagaan lake. Among all of the measured groundwaters and springs around the Böön Tsagaan lake, water chemistry with Cl − /SO 4 2− > 1 is achieved for the groundwater of group B−east (e.g., TN 25 and TN 29) (Figure 7). The relative abundances of Cl − and Na + of group−B−east groundwater are higher than those of the other regions ( Figure 7). Evaporation of the groundwater of TN 29, in fact, results in high Cl − /SO 4 2− in the evaporating water ( Figure 7). Based on the above−mentioned behaviors of the evaporating river and groundwaters, together with our trial−and−error calculations, we can reproduce chemical compositions similar to that of the Böön Tsagaan lake through evaporative concentration of a hypothesized source water (Figure 7). A typical result of the water chemistry evolutional track of a hypothesized source water is shown in Figure 7. The hypothesized water chemistry falls within the range that can be explained by mixing of the river water and groundwaters. In particular, to achieve a chemical composition similar to the lake, the hypothesized source water needs to be close to the mixing line of group−A and group−B−east groundwaters in the piper diagram ( Figure 7). The needs of the groundwater of group B−east suggest that the Böön Tsagaan lake is sustained not only by local seepage of the groundwater of group A, but also by regional flows of the groundwater of group B−east from the areas near Tsagaan Nuur wetland.
Although we do not calculate the evaporative concentrations in detail due to the small number of the measured data, similar trends of evaporative concentration of water chemistry are expected for the Orog and Olgoy lake systems. In order to achieve the lake water chemistry of m Na > (m Cl + 2m SO42 ) for the Orog and Olgoy lakes (Figure 4), the initial source water also needs to be in the lower−left quadrant in the piper diagram, where m Na > (m Cl + 2m SO42 ). Namely, the groundwater of group A could explain the lake compositions due to evaporative concentration. The Tuin river and spring of the Orog lake are candidates for the major water source because their water chemistry is m Na > (m Cl + 2m SO42 ) (group A) ( Figure 4). As for the Olgoy lake, the groundwaters in north of the lake (e.g., the well of Olgoy spring 1 and Gault spring) exhibit the water chemistry of group A with m Na > (m Cl + 2m SO42 ). These results suggest that evaporations of both the river water and/or groundwater would be important to explain the compositions of the Orog and Olgoy lakes. Unlike the Böön Tsagaan lake, the contributions of the rivers may be significant for the Orog and Olgoy lakes. This could be because the Böön Tsagaan lake is larger than the other lakes. For instance, the depth of the Böön Tsagaan lake is~10 m, whereas the depths of the Orog and Olgoy lakes are only 1-3 m. Thus, groundwater could upwell more easily into the deep topographic depression of the Böön Tsagaan lake cavity than the Orog and Olgoy lakes.

Possible Source of Groundwater and Hydrogeochemical Cycles
In Section 5.1.1., we suggest the need for inflows of multiple sources of groundwaters (e.g., groups A and B−east) to explain the water chemistry of the Böön Tsagaan lake; however, how can the different types of water chemistry be generated in the subsurface? Figure 7 shows that a hypothetical water located close to group A on the mixing line between groups A and B−east could explain the lake water composition. This implies that a major water source of the Böön Tsagaan lake would be local seepage of group−A groundwater. In general, groundwater with a water table higher than a lake tends to supply it through local groundwater seepage, e.g., [38] ( Table 1).
Our conclusion that group−A groundwater is a major water source may be inconsistent with previous results suggesting that the Khangai mountains are the ultimate water source due to the aqueous organic compositions of the Böön Tsagaan lake largely being influenced by the occurrence of permafrost in the mountains [27]. The previous study showed that organic compounds bound in water of the Böön Tsagaan lake are derived from melting water of permafrost in the Khangai mountains [27]. To explain the two conflicting results that the lake composition derives largely from group−A groundwater and yet the lake's organic composition is influenced by permafrost thaw in the Khangai mountains, we suggest that the source of the groundwater of group A would be river water that infiltrated the subsurface. This can happen because the groundwater of group A is distributed downstream of the Baidrag river near the Böön Tsagaan lake (Figure 2b). During the river water's infiltration into the subsurface, the water would interact with soil materials. We found carbonate concretions in subsurface soils south and east of the Böön Tsagaan lake (Section 4.2) (Supplementary Figure S1). Dissolution of calcite in the soils would increase the concentrations of Ca 2+ , HCO 3 − , and CO 3 2− , possibly forming a Ca−HCO 3 type (i.e., group A) groundwater. The calcite concretions may have been generated through early diagenesis of carbonate−containing sediments of a larger lake in the past [18]. As described above in Section 1, the Valley of the Gobi Lakes would have been once covered with a larger lake in more humid past [28]. We found bedding of the lacustrine sediments in the outcrops (Supplementary Figure S1). We suggest that early diagenesis of the carbonate−containing sediments might be a source of calcite concretion in the soils. In Section 5.1.1, we suggest the necessity of group−B−east groundwater flowing into the Böön Tsagaan lake. The groundwater of group B−east is distributed around Tsagaan Nuur wetland (Figure 2b). In Figures 4 and 7, we find that the groundwater of group B−east can be generated through a mixing of Ca−HCO 3 −type groundwater with the highly saline shallow pool of Tsagaan Nuur wetland. This shallow pool exhibits extremely high levels of Na + , Cl − , Mg 2+ , and SO 4 2− (Table 3). The near zero saturation indexes of gypsum and calcite suggest that the Ca 2+ , Mg 2+ , HCO 3 − , and SO 4 2− are controlled by dissolution of these secondary minerals in Tsagaan Nuur wetland, whereas the water chemistry of Tsagaan Nuur wetland is unsaturated with epsomite or mirabilite (e.g., the saturation indexes of epsomite and mirabilite are −1.62 and −1.48, respectively, at 25 • C). In semi−arid climates, the water of shallow pools tends to infiltrate the subsurface when its water level is higher than the surrounding groundwater table, e.g., [38]. In fact, the elevation of Tsagaan Nuur wetland (~1285 m) is comparable to or higher than the water table elevations of the group B−east wells of groundwater (1280-1285 m) (Table 1), which suggests that Tsagaan Nuur wetland's surface saline water may have infiltrated the subsurface and formed group−B−east groundwater through mixing. The groundwater of group B−west is located in between the groundwaters of groups A and B−east (Figure 2b). The group−B−west groundwater contains more SO 4 2− and Mg 2+ than group B−east ( Figure 4). Figure 7 shows that the evolution track of the evaporating river water is consistent with the measured compositions of the group−B−west groundwater, suggesting the occurrence of groundwater evaporation of river water that infiltrated soils in this area. In fact, groundwater evaporation occurs at a depth of 1-5 m from the surface depending on vegetation and climate, e.g., [39]. We suggest that the groundwater of group B−west is generated through groundwater evaporation. Figure 8 summarizes a conceptual image of hydrogeochemical cycles and interactions between the surface and groundwater around the Böön Tsagaan lake suggested by the present study. The Baidrag river water not only provides water to the lake, but also infiltrates calcite−containing soils downstream, generating the Ca−HCO 3 −type groundwater of group A near the lake (Figure 8). Local seepage of group−A groundwater would play a central role in supplying water to the lake. In addition to group−A groundwater, a supply of group−B−east groundwater would be important to explain the lake water composition. The groundwater of group B−east would be generated through recharge of highly saline surface water into the subsurface near Tsagaan Nuur wetland (Figure 8), whose water source includes small transient rivers from the mountain area and flooding of the Baidrag river in rainy seasons. The groundwater of group B−east is then supplied into the lake through regional groundwater flows (Figure 8). The driving force of regional groundwater flows into the lake may be owing to the fact that the Böön Tsagaan lake is a deep lake in this area. In addition, regional groundwater flows may be enhanced by effective evaporation of the lake water in semi−arid climates [11], similar to other groundwater−fed lakes that developed in (semi−)arid climates, e.g., [38]. We suggest that the multiple hydrogeochemical processes (i.e., infiltration of surface water, seepage of local groundwater, and regional groundwater flows) would sustain the Böön Tsagaan lake system. Minerals 2020, 10, x FOR PEER REVIEW 23 of 28 [11], similar to other groundwater−fed lakes that developed in (semi−)arid climates, e.g., [38]. We suggest that the multiple hydrogeochemical processes (i.e., infiltration of surface water, seepage of local groundwater, and regional groundwater flows) would sustain the Bӧӧn Tsagaan lake system.

Implications for Hydrogeochemical Cycles around Lakes on Early Mars
Comparing to other terrestrial analogues for early Martian lakes (e.g., saline lakes in the Qaidam Basin and Atacama−Andes regions, and cold springs on Axel Heiberg Island) [22][23][24], the sources of water and salts differ each other depending on the geohydrological settings. Owing to the tectonically−active setting and hyperarid climate of Atacama−Andes regions, there are multiple sources of salts, including aeolian inputs (sea salts and desert dust), recycling of crustal salts, and volcanic inputs [24]. In the polar zone of Axel Heiberg Island springs [22], the major source of highly saline water is subglacial meltwater. Although these supply processes of salts could have also occurred in lake systems on early Mars, the characteristics of the Valley of the Gobi Lakes allow us to understand the roles of evaporative concentration of salinity and groundwater hydrology in semi−arid climates. Despite of the variety of the water sources among the analogue sites, all of them suggest the importance of groundwater hydrology to transport the water and dissolved elements to the lakes.
In Section 1, we raise some questions regarding hydrogeochemical cycles that developed in semi−arid climates on early Mars-(1) Did groundwater flow into lakes? (2) If so, what types of dissolved species were supplied to lakes? (3) Could surface water and salts infiltrate the subsurface? Our field survey for a terrestrial analog of hyposaline lakes in semi−arid climates provides insights into these questions.
Concerning questions (1) and (2), our results suggest that groundwater inflows play central roles in the hydrogeochemistry of the closed−basin lake systems of the Valley of the Gobi Lakes. River water originating from the melting of permafrost and snow in the mountain areas (i.e., the Khangai mountains) would be a major source of water sustaining the lakes. Thus, the melting of permafrost and snow is the ultimate crucial source of water in this area. However, interactions with soils significantly affect the chemical compositions of water that is supplied to the lakes. In the Valley of the Gobi Lakes, calcite in soils would dissolve into the groundwater near the inlet river, providing Ca and HCO3 to the lakes via local groundwater seepages (Figures 3 and 4). On current Mars, surface soils contain high abundances of sulfate minerals (e.g., anhydrite) at the landing sites of Mars rovers and landers, e.g., [40,41]. If sulfate minerals were also contained in soils on early Mars, interactions of infiltrated river water with soils would have provided dissolved Ca 2+ and SO4 2− to the groundwater

Implications for Hydrogeochemical Cycles around Lakes on Early Mars
Comparing to other terrestrial analogues for early Martian lakes (e.g., saline lakes in the Qaidam Basin and Atacama−Andes regions, and cold springs on Axel Heiberg Island) [22][23][24], the sources of water and salts differ each other depending on the geohydrological settings. Owing to the tectonically−active setting and hyperarid climate of Atacama−Andes regions, there are multiple sources of salts, including aeolian inputs (sea salts and desert dust), recycling of crustal salts, and volcanic inputs [24]. In the polar zone of Axel Heiberg Island springs [22], the major source of highly saline water is subglacial meltwater. Although these supply processes of salts could have also occurred in lake systems on early Mars, the characteristics of the Valley of the Gobi Lakes allow us to understand the roles of evaporative concentration of salinity and groundwater hydrology in semi−arid climates. Despite of the variety of the water sources among the analogue sites, all of them suggest the importance of groundwater hydrology to transport the water and dissolved elements to the lakes.
In Section 1, we raise some questions regarding hydrogeochemical cycles that developed in semi−arid climates on early Mars-(1) Did groundwater flow into lakes? (2) If so, what types of dissolved species were supplied to lakes? (3) Could surface water and salts infiltrate the subsurface? Our field survey for a terrestrial analog of hyposaline lakes in semi−arid climates provides insights into these questions.
Concerning questions (1) and (2), our results suggest that groundwater inflows play central roles in the hydrogeochemistry of the closed−basin lake systems of the Valley of the Gobi Lakes. River water originating from the melting of permafrost and snow in the mountain areas (i.e., the Khangai mountains) would be a major source of water sustaining the lakes. Thus, the melting of permafrost and snow is the ultimate crucial source of water in this area. However, interactions with soils significantly affect the chemical compositions of water that is supplied to the lakes. In the Valley of the Gobi Lakes, calcite in soils would dissolve into the groundwater near the inlet river, providing Ca and HCO 3 to the lakes via local groundwater seepages (Figures 3 and 4). On current Mars, surface soils contain high abundances of sulfate minerals (e.g., anhydrite) at the landing sites of Mars rovers and landers, e.g., [40,41]. If sulfate minerals were also contained in soils on early Mars, interactions of infiltrated river water with soils would have provided dissolved Ca 2+ and SO 4 2− to the groundwater downstream of valley networks and channels, which could have eventually supplied Ca 2+ and SO 4 2− to lacustrine environments on early Mars. In the lacustrine mudstone−sandstone of the Sutton Island member of the Murray formation of Gale Crater, sulfate minerals are thought to exist in the matrix of the sediments, suggesting precipitation of lacustrine sulfate [16]. We suggest that local groundwater seepage could have been one of the sources of dissolved sulfate in the early Gale lakes if sulfate minerals existed in the soils at the time of deposition, although the direct inputs of sulfate from the atmosphere and surface cannot be excluded. In early lakes in Jezero Crater, carbonate minerals (e.g., magnesite) may have been present in the soils downstream of inlet rivers and alluvial fans, given the presence of carbonate minerals (Mg carbonates) in the catchment area of valley networks of Jezero Crater (i.e., Nili Fossae) [42]. Our results suggest that if the soils around Jezero Crater contain detrital Mg carbonates, groundwater beneath inlet rivers and alluvial fans would have contained an abundance of dissolved Mg 2+ and HCO 3 − .
Local seepages of the groundwater would have then supplied these dissolved species to the lake. The supplied Mg 2+ and HCO 3 − would have been consumed within the early Jezero lakes through evaporation and consequent precipitation of authigenic carbonate minerals, such as AMC, in a similar manner to the terrestrial saline lakes found in the Valley of the Gobi Lakes [35,36]. In fact, recent remote−sensing observations found possible evidence for lacustrine carbonates (hydrated magnesite and/or hydromagnesite) near a large inlet river and delta deposits of Jezero Crater [43]. Fukushi et al. (2017) show that AMC is the precursor of deposition of hydromagnesite [37]. Given the low degree of diagenesis and preservations of metastable phases (e.g., ferrihydrite) in lacustrine sediments of Gale Crater, e.g., [38], we predict that metastable AMC and/or diagenetic hydromagnesite could be preserved in the lacustrine sediments of Jezero Crater. The findings of AMC/hydromagnesite can serve as evidence for authigenic precipitation of carbonate minerals in alkaline lakes [35,43], which may be testable by the Mars 2020 Perseverance rover and in future Mars Sample Return missions. Regarding question (3) of the fate of surface salts, our results suggest that saline shallow pools in the wetlands (i.e., Tsagaan Nuur wetland) infiltrate the subsurface, providing Cl − and Na + to the groundwater (Section 5.1). On early Mars, shallow pools in wetlands might not only have contained Cl − and Na + , but also photochemically produced oxychlorides, such as ClO 3 − and ClO 4 − [44].
Irradiation from solar UV light would have also oxidized and acidified surface water through photo−oxidation of ferrous ions [45]. If oxidizing surface water infiltrated the wetlands of semi−arid climates on early Mars, highly reactive oxychlorides (e.g., ClO 3 − ) would have oxidized the subsurface materials, as suggested previously [44,46]. Less reactive Cl − , Na + , and ClO 4 − in infiltrating surface water could have been transported into deep lakes through regional groundwater flows, as proposed by [8]. The previous study estimated a duration of hydrological cycles based on observed thickness of salt deposits, assuming Na + and Cl − were supplied by water-rock reactions using a reactive transport model [8]. The accumulation of pre−existing surface salts into lakes through regional flows, however, could affect the estimation of the duration [8]. Finally, transport of surface ClO 4 − through regional groundwater flows to lakes could have contributed to generate redox disequilibria in the lacustrine environments. On Earth, similar transports of oxidants (e.g., NO 3 − ) via regional groundwater flows and reductants (e.g., Fe 2+ and Mn 2+ ) via upwelling of deep groundwater controls microbial ecosystems in a floodplain in semi−arid climates [47]. Together with upwelling of reducing deep groundwater in semi−arid climates on early Mars [11], such transports of multiple sources of groundwater could provide a perspective on the past habitable environments on Mars.

Conclusions
The present study focused on the reconstruction of the geohydrological cycles of the closed−basin lake systems in the Valley of the Gobi Lakes, central Mongolia, based on field surveys. The water of the Valley of the Gobi Lakes largely relies on the melting of snow and permafrost in the mountain ranges. The melted water is transported to closed−basin lakes in semi−arid steppe plains where evaporation predominates precipitation. This hydrological nature would allow this area to be a promising terrestrial analog of lacustrine environments that developed in cold and semi−arid climates on early Mars. Since the sampling locations are still sparse compared with the catchment area, especially in the west shore area of the Böön Tsagaan lake, the whole picture of hydrology and geochemical cycles of the lake systems remains unclear. Additionally, the measured water chemistry data may be snapshots of seasonal or long−term variations. Despite these uncertainties, we can summarize the following results of our field study.
(1) The water chemistry of the closed−basin lakes is characterized as alkaline−hyposaline, Na−Cl−type water. This is in contrast to the circumneutral, low−salinity waters of the rivers and groundwaters. Even for the groundwater adjacent to the lakes, salinity remains at low levels, suggesting ineffective outflowing of the lake water to the surrounding subsurface. We suggest that groundwater predominantly inflows into the lakes in semi−arid climates, which would be driven by effective evaporation of surface water.
(2) The groundwater chemistry around the lakes can be classified into a few groups. In group A, water chemistry is characterized as Ca−HCO 3 type, which would be generated by interactions of infiltrating river water with calcite−containing soils near the inlet rivers. In group B−east, groundwater contains Cl − and Na + together with Ca 2+ , Mg 2+ , HCO 3 − , and SO 4 2− , which would be formed by the mixing of infiltrating surface brine with groundwater. The high−salinity brine occurs in shallow pools of the wetland (Tsagaan Nuur wetland) upon evaporation. In group B−west, the groundwater contains Ca 2+ , Mg 2+ , and SO 4 2− , which would be generated through groundwater evaporation.
(3) The lake water composition cannot be explained solely by evaporation of the river water. To achieve the composition of the Böön Tsagaan lake, inflows of the different types of groundwaters are necessary. In particular, the groundwater of groups A and B−east would be important to account for the lake composition. These results suggest that the occurrence of local seepage (group−A groundwater) and regional flows (group−B−east groundwater) are required to sustain the Böön Tsagaan lake.
These results would provide insights into hydrogeochemical cycles around closed−basin paleolakes on Mars. In semi−arid climates of early Mars, groundwater may have played central roles not only in sustaining surface water in lakes, but also in collecting soluble species in the subsurface of the catchment area. For instance, local groundwater seepage could have been a source of Ca 2+ , Mg 2+ , and SO 4 2− to lakes through dissolution of sulfates in soils. If the soils near early Jezero lakes contained detrital Mg carbonates provided from the Nili Fossae region, local groundwater seepage would have also provided an abundance of dissolved Mg and alkalinity to the lakes, which in turn could have caused authigenic precipitation of lacustrine carbonates [43]. Our results also support the idea that shallow pools in wetlands in semi−arid climates of early Mars would have transported salinity and oxidizing materials to the subsurface groundwater through infiltration [8,44]. We suggest that closed−basin lakes in semi−arid climates on early Mars would have been sustained by multiple types of groundwater inflows (e.g., local seepage, regional flows, and upwelling of deep groundwater [11]), which would have generated redox disequilibria in the lacustrine environments and could have affected the past habitability.