There are many lakes in the Qinghai region and Qaidam Basin. Most of these lakes are classified as saline lakes, in that they possess a salinity with the variation from 8.86 to 72.36‰. The water samples used in this study were collected from six lakes with different salinities. Four of the sampled lakes were located in the northeastern part of the Qaidam Basin, including Gahai Lake (GH), Keluke Lake (KLK), Tuosu Lake (TS), and Xiligou Lake (XLG) (Figure 1). These four lakes can be divided into three categories based on the basis of salinity: freshwater lakes (including KLK), saline lakes (TS, XLG) and salt lakes (GH). The study area is characterized by a typical alpine continental climate in which large temperature differences occur between day and night. Precipitation in the region is low, while evaporation throughout the year is relatively high. The main vegetation types include alpine meadows and alpine step communities. The overall elevation of the area is about 3,000 m masl. Han et al. (2019) determined, using data from the National Meteorological Station of Delhi City, that the mean annual precipitation over the period of 1961–2017 was 185.1 mm, and the mean annual evaporation ranged between 2,242.8 and 2,439.4 mm. Precipitation is unevenly distributed during the year, showing a unimodal distribution. Maximum precipitation occurs in June to July, while the minimum precipitation occurs in December. The total area of GH is about 37 km²; it possesses a mean water depth of 8 m and a maximum water depth of 15 m. The lake basin has no perennial surface rivers. Recharge relies on atmospheric precipitation and subsurface runoff. TS has a total area of about 180 km² and a maximum water depth of 25.7 m. The main source of the water in TS originates from drainage, atmospheric precipitation, and groundwater (Zheng and Zhang, 2002; Huang and Han, 2007).

In addition to the four lakes sampled in the Qaidam Basin, water and ice samples were collected from two lakes with varied salinities in the Qinghai region: Qinghai Lake (QHH) and Qinghai Gahai (QHGH) (Figure 1). The surface area of Lake Qinghai (36°15′–38°20′N, 97°50′–101°20′E) is approximately 4,400 km². The area of the lake is approximately 29,660 km². The geological and geographical environment and current climate of this area have been examined in detail by Colman et al. (2007) and Jin et al. (2010).

Lake Qinghai is the largest inland close lake in China by surface area. The lake is developed within a basin surrounded by three mountain ranges including the Datong Mountains to the north, the Riyue Mountains to the east, and the Qinghai Nanshan Mountains to the south (Bian et al., 2000). The lake is divided into two nearly equal sub-basins by a NNW-trending horst, from which an island (Mt Haixin) emerges. Several minor fault scarps also can be found in the lake basin. The lake bed is generally flat and the lake is hydrologically closed and evaporative (Zhang et al., 1994). The six largest rivers that form the Lake Qinghai catchment are, in order of their discharge, the Buha, Shaliu, Hargai, Quanji, Daotang and Heima rivers. Hydrologically, these rivers are greatly impacted by the area’s monsoonal climate, and they supply over 87% of the water, dissolved load, and sediment discharge to the lake. Over half of these rivers come from the Buha River (Jin et al., 2009). Meltwater from surrounding mountain glaciers accounts for only 0.3% of the total runoff (Colman et al., 2007).

Qinghai Gahai Lake (36°57′–37°3′N, 100°31′–100°36′E) is located along the northeast bank of Qinghai Lake at an altitude of 3,196.6 m. As one of the three sub-lakes of Qinghai Lake, Qinghai Gahai is now hydrologically closed without river inputs. The lake is 12 km long from northwest to southeast, and 6 km wide from southwest to northeast. It possesses a surface area of 47.5 km². Qinghai Gahai Lake is generally 8–9.5 m deep and slightly alkaline with a pH of 9.25. The salt content of the lake water is about 31.73 g/L. The lake is recharged by shallow groundwater. The center of the lake is located about 400 m from the coast, and there is deep groundwater recharge at the bottom of the lake.

All ice and water samples were collected in February 2021 (see Table 1 for details). To determine the effect of collecting samples in different years on U content in lake water, four samples were collected in Qinghai Lake from 2017 to 2019. In addition, river water and groundwater samples were collected from the rivers recharging Qinghai Lake to verify whether surface runoff would affect lake U concentrations and [²³⁴U/²³⁸U]_AR. Previous studies showed that the U concentration and [²³⁴U/²³⁸U]_AR of waters from the same lake in the Qaidam Basin varied little aerially or vertically within the water column or seasonally through time (Zhang et al., 2019; Zhao et al., 2020). Thus, we believe that the samples from one location of each lake are sufficient to represent the U content and the [²³⁴U/²³⁸U]_AR distribution characteristics of these sampled lakes, and that the U data of surface water can represent the entire water column of the lake. Sampling periods were selected that possessed large quantities of lake ice in the northeastern part of the Qaidam Basin and the Qinghai region; during the other seasons, the lakes are nearly ice free. Thus, the sampled ice and lake waters should show the typical U distribution within the ice and the underlying water. By comparing the U composition of the sampled lake water in summer with the maximum water input published in Zhao et al. (2020), we expected to gain a deeper understanding of effects of ice freeze-thaw processes on U isotope distribution in saline lake waters and the potential environmental implications on the observed U data.

To further evaluate the effects of ice freeze-thaw processes on the U isotope distribution of lake water, three water samples were collected from QHH in September 2018, July 2019 and February 2021 (Table 2). We preformed inhouse freezing experiments on these three QHH water samples in the laboratory to further verify the results of the data obtained from the “natural” ice and water samples collected in the field.

Samples of surface ice and the underlying water were collected dozens of meters offshore in high density polyethylene (HDPE) plastic bottles (Nalgene 2002-0032) by hand wearing Derma Free Vinyl Gloves. After collection, about 1,000 ml of each water sample was filtered through a mixed cellulose filter membrane (0.8-µm pore size, Millipore AAWP04700) with a manual vacuum filter system (Nalgene, 300-4100; Nalgene, 6133-0010). The filtrates were stored for later analysis.

After collection, ice samples were placed at room temperature to melt. The melted ice water was also filtered through a mixed cellulose filter membrane (0.8-μm pore size, Millipore AAWP04700) using a manual vacuum filter system (Nalgene, 300-4100; Nalgene, 6133-0010), and stored for later analysis. The research of Ku et al. (1977) showed that there is no evident difference on the [²³⁴U/²³⁸U]_AR on the filtered Atlantic seawater and the unfiltered seawater, but has little effect on U concentration. Therefore, we believe that U concentration filtered by 0.8 μm filters could represent the content of dissolved U in saline lake water to a certain extent.

QHH water samples were collected in three batches over various sampling periods for in-house freezing experiments. The experiment mimicked natural sample freezing conditions in the field, during which the three previously filtered QHH water samples were frozen at −18°C in a freezer for 5–6 h to form half water and half ice. Then the ice and water were separated, and stored for later analysis.

The samples were analyzed for selected water chemistry parameters (e.g., anions, cations, salinity, total dissolved solids (TDS), and pH at the Key Laboratory of Surface System and Environmental Carrying Capacity of Shaanxi Province, Northwest University. Anion and cation concentrations of the samples were determined using two Dionex AQUION ICs (Chen et al., 2007; Scientific, 2016). Conductivity (ms/cm), TDS (g/L), salinity (‰), and pH were determined using a Mettler-Toledo Five Easy Plus pH meter. Uranium isotopes in the samples were measured at the Xi’an Jiaotong University Isotope Laboratory. The analysis of U isotope abundances included a two-step process, including chemical separation and instrumental analysis. Chemical separation involved sample digestion, centrifugal coprecipitation, ion exchange resin separation, uranium isotope collection, and purification. More specifically, to analyze the U isotopic compositions of the water, the acidified aliquots of each sample (∼25 ml) were spiked with ²³³U–²³⁶U–²²⁹Th, preconcentrated and digested with HNO₃ and HClO₄, and then co-precipitated with Fe oxyhydroxide. The Fe precipitate was moved to a centrifuge tube for centrifugation and rinsed with deionized H₂O (>18 MΩ) to eliminate water-soluble seawater cations. The precipitate was then dissolved in 14 N HNO₃ (1 N = 1 M) and moved to a Teflon beaker, before being dried and dissolved in 7 N HNO₃ for anion-exchange chromatography with an AG1-X8, 100–200 mesh size resin and a polyethylene frit. Separation was performed in Teflon columns with a 0.25 ml column volume (CV). The first separation was performed in Teflon columns with 0.5 ml 7 N HNO₃, 0.25 ml of 8 N HCl (to get rid of Fe and Th fraction), and 0.5 ml of deionized H₂O (to gather the U fraction). The U fractions were then dried with two drops of HClO₄ and brought to volume with 7 N HNO₃. The final U fractions were then dried with two drops of HClO₄ and dissolved in weak nitric acid for analysis on the mass spectrometer. The concentrations of ²³⁴U, ²³⁵U, and ²³⁸U were calculated by isotope dilution using the nuclide ratios determined on a Thermo-Finnigan Neptune mass spectrometer. All measurements were performed using a peak-jumping routine in ion counting mode on the discreet dynode multiplier behind the retarding potential quadrupole. Each sample measurement was bracketed by the measurement of an aliquot of the run solution, which was utilized to adjust for the instrument background count rates on the measured masses. For U, the measurement uncertainties involved propagated errors from the ICP-MS isotope ratio measurements, spike concentrations, and blank corrections. The procedural blanks for chemical and mass spectrometric analyses at the Laboratory of Isotope Geochemistry in Xi’an Jiao Tong University are approximately 50 ag (1.3 × e⁵ atoms) for ²³⁴U; 3 fg (7.7 × e⁶ atoms) for ²³⁵U; and 150 fg (3.8 × e⁸ atoms) for ²³⁸U.

The experimental analysis methods can be divided into two parts: chemical separation and MC-ICP-MS mass spectrometry. For chemical separation, we used duplicate samples in each batch to ensure the reliability of the data. The experimental results showed that the repeatability of the duplicate samples in the chemical analysis process was excellent (Table 3), and the yield of all samples was about 93%. For instrument analysis process, NBL-112A (New Brunswick Laboratory) U isotope standard solution was used for external calibration during the analysis by MC-ICP-MS mass spectrometer. The experimental results showed that the average of NBL-112A standard solution was −38.23 ± 1.21‰(2σ), and the measurement precision of dissolved ²³⁸U content and [²³⁴U/²³⁸U]_AR in lake water samples was 0.89–1.35‰ (2σ). The methods used herein have been fully described by Cheng et al. (2000, 2013) and Shen et al. (2002, 2012).

The water chemistry results of each sampled lake are presented in Table 4, including the aqueous ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR. Aqueous ²³⁸U concentrations of the freshwater lakes in the northeastern part of the Qaidam Basin were generally low, which is consistent with the results of previous analyses of lake water samples collected in June, July and the beginning of December without ice coverage (Zhao et al., 2020). The ²³⁸U concentration of ice from the five sampled lakes were less than that of water under the ice, except for samples from QHH. The ²³⁸U concentrations of the ice from the five lakes with varying salinity ranged from approximately three to forty percent of the U content of the underlying water. The lower the salinity of lake water, the lower U content in the forming overlying ice. The ²³⁸U concentration of ice in QHH was slightly less than that of water under the ice. The reason for this is not clear at this time and will be the topic of future studies.

The [²³⁴U/²³⁸U]_AR values were similar between the ice and water from the six lakes of varying salinity; the [²³⁴U/²³⁸U]_AR exhibited very limited variation. In fact, the variation in [²³⁴U/²³⁸U]_AR ratios between the ice and water in the five saline lakes and the salt lake were all within 2%. The variation in [²³⁴U/²³⁸U]_AR between ice and water in the freshwater lake reached up to 23%.

In conclusion, during the icing process, the U content was significantly different between the ice and water; it was lower in ice than in water. In contrast, the [²³⁴U/²³⁸U]_AR did not change significantly between the saline lakes; differences in the values were within 2% and slightly lower in ice than the underlying water.

Laboratory studies were carried out to 1) further determine the fidelity of the isotopic changes observed during the freezing and transformation of “natural” lake water to ice, and 2) assess the anomalies in the results of ice and water in Qinghai Lake during the ice-on (covered) season. The indoor freeze-thaw experiments were conducted on three batches of QHH water samples collected in different seasons, including periods when the lake was and was not covered in ice. The samples from QHH were obtained in September 2018, July 2019 and February 2021.

The results of the inhouse freezing experiments were similar to that observed for the natural lake samples (Table 5). The ²³⁸U concentration of the ice was slightly less than that of the water in all three batches of samples. The ²³⁸U concentration of the ice was about 40% of that in the water. In addition, we noticed that the ²³⁸U concentrations of the water collected in February were lower than those of the water obtained in July and September. The [²³⁴U/²³⁸U]_AR resulting from the inhouse freezing experiments also supports those of the nature samples. The activity ratios did not change significantly during the phase transition from water to ice in the three batches of collected samples; differences between the samples were within 0.5%.

In November 2020, five groundwater samples in recharge rivers and three water samples from recharging rivers into Qinghai Lake. The water chemistry results of each sampled river are presented in Table 6, including the aqueous ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR. Aqueous ²³⁸U concentrations of the recharge rivers of Qinghai Lake is 2.1–2.6 μg/L, and [²³⁴U/²³⁸U]_AR is 2.010–2.325, with a limited variation. The ²³⁸U concentrations and [²³⁴U/²³⁸U]_AR of the groundwater in Qinghai Lake region ranged from 3.8 to 14.4 μg/L and 1.442 to 2.562 respectively. The U concentration of recharge rivers and groundwater is lower than that of Qinghai Lake, while [²³⁴U/²³⁸U]_AR is higher than that of Qinghai Lake.

Global greenhouse gas warming controlled erosion processes are predicted to affect 61% of the ice-free landmasses during the 21st century, a number that increases to 80% when land use is considered (Ostberg et al., 2018). These processes have the potential to induce large variability in the weathering of the continents and the composition of chemical fluxes to the ocean. Seawater [²³⁴U/²³⁸U]_AR provide global-scale information about continental weathering and are vital for marine uranium-series geochronology. Existing research suggests that the Atlantic [²³⁴U/²³⁸U]_AR started to increase before major sea-level rise and overshot the modern value by 3‰ during early deglaciation. In addition [²³⁴U/²³⁸U]_AR during deglacial times in the Pacific Ocean converged with that in the Atlantic after the abrupt resumption of Atlantic meridional overturning circulation. This led Chen et al. (2016) to suggest that ocean mixing and early deglaciation released excess ²³⁴U through the enhanced subglacial melting of the Northern Hemisphere ice sheets, and the change in ²³⁴U has driven the observed evolution in [²³⁴U/²³⁸U]_AR. To assess the reliability of this hypothesis, it is necessary to evaluate the migration and transformation of ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR during the transformation between ice and seawater of a given salinity and set of hydrochemical conditions. Therefore, we conducted this study to explore whether the freezing and thawing process of ice affects the U concentration and [²³⁴U/²³⁸U]_AR in lake waters, which were easier to sample than waters in polar regions.

As shown in Figure 2, the ²³⁸U concentration of the ice from the five lakes with varying salinity was approximately three to forty percent of the U content of the overlying water. Moreover, the lower the salinity, the lower the ratio of U in ice to the underlying water. The [²³⁴U/²³⁸U]_AR between ice and water in the six lakes with varying salinity were similar and varied within a narrow range of about 2%. All exhibited lower [²³⁴U/²³⁸U]_AR in ice than in the underlying water. In addition, the [²³⁴U/²³⁸U]_AR observed during the inhouse freezing experiments of water from QHH supported the observations of the nature lake samples; differences in the ratios between ice and water in the three batches of samples did not change significantly as they were within the measurement error (Figure 3). In conclusion, the ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR resulting from the indoor freeze-thaw experiments support the results of the field observations of the natural saline lake samples. The concentration of ²³⁸U of seawater consistently varies by about 3.2 ppb between the Pacific and Atlantic Oceans (Chen et al., 1986). The ²³⁸U concentration of ice possessed a lower concentration than that of water, and thus did not result in an observable change in the ²³⁸U concentration of seawater.

These data show that the freezing and thawing process of ice does not cause [²³⁴U/²³⁸U]_AR to change in the remaining water. Chen et al. (2016) did mention that other factors other than ice melting could be responsible for the increase [²³⁴U/²³⁸U]_AR. However, their main point was that the early deglacial release of excess ²³⁴U drove the observed evolution in [²³⁴U/²³⁸U]_AR. If the increase of [²³⁴U/²³⁸U]_AR was dependent on factors other than ice melting (e.g. the leaching of continental matter into the water), then the Pacific and Atlantic oceans at lower latitudes would have warmed faster, melted faster, and responded faster to the [²³⁴U/²³⁸U]_AR change, which is not that the case.

To try to understand the cause of the above phenomenon, we first explored whether the freeze-thaw process of ice leads to an increase in the activity ratio of the water column using natural samples and inhouse freeze-thaw experiments. The preliminary results showed that the freezing and thawing process of ice does not cause [²³⁴U/²³⁸U]_AR increase of the remaining water. Second, we explored whether the input of the river could increase the [²³⁴U/²³⁸U]_AR of the remaining water. Current estimations of the average world river uranium concentration (0.3–0.6 µgl⁻¹) and [²³⁴U/²³⁸U]_AR (1.2–1.3), combined with the diffusional ²³⁴U influx from sediments (0.3 dpm cm⁻² 10⁻³yr⁻¹), are essentially consistent with a model that depicts a steady state distribution of uranium in the ocean. However, the 0.3–0.6 µgl⁻¹ values for river uranium may be an upper limit estimate (Ku et al., 1977). The ratios reported for global rivers ([²³⁴U/²³⁸U]_AR = 1.171) (Chabaux et al., 2008) and the ratios of four rivers discharging into the Lake Qinghai catchment located in an arid alpine climate possessed [²³⁴U/²³⁸U]_AR varying from 1.984 to 2.506 (Zhang et al., 2019). New Zealand rivers located in alpine high latitude regions exhibit a very wide range of [²³⁴U/²³⁸U]_AR, varying between 1.09 and 4.61 (Robinson et al., 2004). This range spans that previously measured in rivers from other locations. Based on the above [²³⁴U/²³⁸U]_AR in river waters from low latitudes, and the ratios reported for global rivers ([²³⁴U/²³⁸U]_AR = 1.171) (Chabaux et al., 2008), it would be difficult for low latitude riverine inputs to result in a 3 per mil increase in the [²³⁴U/²³⁸U]_AR of sea water in the Pacific and Atlantic because of a higher [²³⁴U/²³⁸U]_AR input of river. Thus, we speculate that the North Atlantic Ocean, which is also located at low latitudes, responds faster than the Pacific Ocean, most likely because the Arctic Ocean, located at high latitudes, carries inputs from five recharging rivers characterized by sandy and clay particle debris, then when the early deglaciation occurs. The inputs from rivers with high [²³⁴U/²³⁸U]_AR leads to an increase in the [²³⁴U/²³⁸U]_AR in the Arctic Ocean, which in turn affects the North Atlantic waters at low latitudes first and then affects the Pacific Ocean by way of the oceanic conveyor belt. Third, uranium is removed from the oceans by diffusion across the sediment-water interface of organic-rich sediments. This pathway is the largest single sink in the global budget of this element (Klinkhammer and Palmer, 1991). Dissolved uranium is drawn into suboxic sediments along a concentration gradient established by the precipitation of an insoluble phase which forms when U(VI) is reduced to U(IV) (Ku et al., 1977). Cowart (1980) measured the concentration of dissolved uranium and [²³⁴U/²³⁸U]_AR were determined in water samples from 23 locations in the Edwards carbonate aquifer of south central Texas, and found that the [²³⁴U/²³⁸U]_AR was about 1.20 in an oxidized aquifer, and above 2.0 in reduced aquifer. Thus, during deglaciation, there exists one possibility, the reduced seawater below the ice is mixed into the overlying seawater.

In summary, we suggest when early deglaciation occurs, the mixing of reduced seawater and/or the inputs from rivers with high [²³⁴U/²³⁸U]_AR leads to an increase in the [²³⁴U/²³⁸U]_AR in the Arctic Ocean, which in turn affects the North Atlantic waters at low latitudes first before influencing the Pacific Ocean by way of the oceanic conveyor belt. The specific reasons for this phenomenon need to be explored in future studies.

Based on the results of the in-house freezing experiments, there were differences in [²³⁴U/²³⁸U]_AR in the ice and water sampled in February (winter) and July/September (summer); lower [²³⁴U/²³⁸U]_AR values were measured in winter than in summer. However, there were no differences in [²³⁴U/²³⁸U]_AR between the ice and water samples obtained in February (winter) (Table 5; Figure 3). Similar results were observed for the natural lake samples collected in winter and summer (Table 4, Table 5).

These results, combined with the limited ice volume in the Arctic, provide insights into the observed change in [²³⁴U/²³⁸U]_AR in the Atlantic, which started to increase before major sea-level rise and overshot the modern value by 3‰ during early deglaciation. We speculate that the phenomenon may be due to a change in the redox conditions to a more oxidized state at the bottom of the water column during the early stages of deglaciation and/or from the inputs of river waters from high altitude regions that usually possess a higher activity ratio. Because at pH values and CO₂ concentrations typical of oxic seawater, the predominant uranium species, U(VI), is highly soluble and occurs as a stable carbonate ion complex, Ca/Mg-UO²⁻(CO₃) complexes (Djogić et al., 1986; Dong and Brooks, 2006; Chen et al., 2020). In contrast, in oxygen-depleted waters, rapid removal of U by the sediment depends on reduction of U(VI) to U(IV) (Tribovillard et al., 2006; Algco and Tribovillard, 2009). In addition, a large quantity of uranium in depositional archives is mostly bound to organic matter as demonstrated by weak to strong positive correlations between naturally occurring uranium and OM contents (Min et al., 2000; Regenspurg et al., 2010; Francke et al., 2020). Uranium can be mineralized in the presence of OM forming strong UO₂-OM bonds (Lenhart et al., 2000; Li et al., 2014). If the uranium bound to residual OM controls [²³⁴U/²³⁸U]_AR, then non-detrital uranium is usually characterized by values above 1. Thus, when polar ice melting is combined with the oxidation of the sea floor, the minerogenic U-OM bounds and/or U-carboxyl groups should be preferentially oxidized and released with higher [²³⁴U/²³⁸U]_AR. This hypothesis needs to be tested by further work.

The accurate evaluation of excess ²³⁴U is important for marine uranium-series geochronology; thus, it is also necessary to investigate the causes of the variation in the [²³⁴U/²³⁸U]_AR of seawater. For example, if the ²³⁴U concentration of coral during glacial periods is about 3.2 ppm, and the initial [²³⁴U/²³⁸U]_AR differs by 3‰, then the determined age will differ by thousands of years (Esat and Yokoyama, 2006).

Previous studies of ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR showed that there was no significant change between sampling sites in the same saline lake based on our published paper (Zhang et al., 2019; Zhao et al., 2020). There is also no evident difference between samples that were collected from a specific lake on different dates during the summer without ice coverage. Lake water samples collected during periods of maximum and minimum discharge have also been used to evaluate the seasonal variations in aqueous ²³⁸U. We found that ²³⁸U concentration did not change with the seasons in these saline lakes. If the ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR remain consistent through time, then sediment ages and/or sedimentation rates can be determined from lake sediment and/or biogenic carbonate in the future, thus allowing for the accurate reconstruction of paleoclimatic and paleoenvironmental conditions.

Since the saline lakes in the Qaidam Basin and Qinghai region have a freezing period of about 3–4 months per year, it is necessary to determine whether the ²³⁸U concentration and the [²³⁴U/²³⁸U]_AR remain consistent. The results obtained herein show that differences in ²³⁸U concentration existed between the ice and the underlying water in all sampled saline lakes and in fresh water. Significant differences in ²³⁸U concentration occurred between winter ice and summer lake waters (without ice) that are significantly recharged by exogenous rivers, including TS, KLK, and QHH. Conversely, there was no significant differences in ²³⁸U concentration between winter ice and summer waters (without ice) in saline lakes receiving less exogenous riverine inputs, including QHGH, XLG and GH [lake water data in summer were published in Zhang et al. (2019) and Zhao et al. (2020)]. Therefore, in the arid Qaidam Basin and Qinghai region, the freeze-thaw process of ice does not appear to affect the concentration of ²³⁸U in saline lakes, but the proportion of exogenously recharged water may affect its concentration. In addition, when ice exists with limited oxygen exchange with atmospheric, with the increase of water column, changes in the redox environment at the sediment-water interface may also affect the concentration of ²³⁸U in the water column. Based on our results, the ²³⁸U concentration in lacustrine sedimentary secondary carbonate minerals and the interannual resolution of biogenic carbonate shells are expected to provide useful insights into lake paleohydrological changes. The ²³⁸U concentration of biogenic carbonate shells may possess a seasonal resolution that could reveal the redox environment at the water-sediment interface in QHH.

There were no significant differences in the [²³⁴U/²³⁸U]_AR of the five saline lakes between ice and the underlying water when both were collected at the same time. The differences in the [²³⁴U/²³⁸U]_AR were all within 2%, with lower [²³⁴U/²³⁸U]_AR in ice. Except for QHGH, there were no significant differences (i.e., exceeding measurement error) in the [²³⁴U/²³⁸U]_AR between winter ice and summer waters on various sampling dates. By contrast, there were significant differences in the [²³⁴U/²³⁸U]_AR of lake waters between winter ice and summer waters in QHH, XLG, GH, TS and KLK (see Zhang et al. 2019; Zhao et al., 2020). The differences in U concentration and [²³⁴U/²³⁸U]_AR in lake water between freezing and non-freezing seasons may be due to the influences of several factors such as surface runoff, groundwater and ice volume.

The study area of this paper is in an arid area of high altitude. Most of the lakes studied are closed saline lakes, and the area of surface runoff and underground recharge of these lakes are very limited. Taking Qinghai Lake as an example (which has the largest proportion of replenishment in the study area), the average water level is 3,193.2 m, the lake surface area is 4,260 km², and the lake water reserve is 71.593 billion m³ (Liu, 2004). Using changes in calculated water volume from 1959 to 2000, the average annual surface water inflow to Qinghai Lake was 1.526 billion m³. Precipitation on the lake surface contributed about 1.561 billion m³ to the lake, and groundwater replenished 603 million m³ of water to the lake (Liu, 2004). The ratio of precipitation over the lake surface to recharge for Qinghai Lake is the highest of the lakes studied. Zhu and Xie (2018) summarized the annual and seasonal precipitation of Qinghai Lake in different years during the past 50 years. The annual precipitation of Qinghai Lake was concentrated in summer and autumn, and the precipitation was the lowest in winter (Zhu and Xie, 2018). In the previous study, we selected samples collected in Qinghai Lake in different years with different precipitation patterns, and the experimental results showed that precipitation and surface runoff in Qinghai Lake had almost no effect on U content and [²³⁴U/²³⁸U]_AR of the lake (Table 7). In 2020, we collected water from a river that recharged Qinghai Lake and groundwater near the recharge river and measured its U content and [²³⁴U/²³⁸U]_AR (Table 6). The results showed that the effects of atmospheric precipitation, surface runoff and groundwater on U concentration and [²³⁴U/²³⁸U]_AR of Qinghai Lake water were negligible. In addition, the samples studied in this paper were collected at the end of February, when the water in Qinghai Lake exhibited its maximum degree of ice cover. The average depth of Qinghai Lake is about 21m, and the average thickness of the ice sheet is about 29.4 cm. Only ∼2% (by volume) of the surface water was frozen. Consequently, the limited volume of ice will not affect U concentration and [²³⁴U/²³⁸U]_AR of the sampled saline lakes.

The ice freeze-thaw processes have almost no effect on the uranium content and [²³⁴U/²³⁸U]_AR of the sampled saline lakes, which were characterized by a limited recharge volume from surface runoff, groundwater, and ice volume, namely the close saline lake in arid alpine background. An exception was U content in Qinghai Lake. We speculate that the low U concentration in the water column of Qinghai Lake during winter freezing is due to the 30 cm thick ice layer that prevents the infiltration of atmospheric oxygen, with the increase of depth of water column (∼21 m). The ice-water interface resembles a redox interface, resulting in a low U content of the water column under the ice.

In summary, the freeze-thaw processes of the lake will not result in significant changes in [²³⁴U/²³⁸U]_AR in excess of 2% in the water column of saline lakes. The [²³⁴U/²³⁸U]_AR in these saline and freshwater lakes are closely related to the characteristics of the input water and the associated water–rock interactions involving sediments, atmospheric dust and organic material, among other factors during the evolution stage, metamorphous degree, and hydrochemistry of the saline lakes. Based on these results, we suggest that there is a potential to maintain relatively consistent levels of [²³⁴U/²³⁸U]_AR in saline lakes characterized by a closed lake basin and which have a limited input of exogenous recharge water and limited organic matter. Usually, the [²³⁴U/²³⁸U]_AR of dissolved water will increase when organic matter is present (Francke et al., 2020). The [²³⁴U/²³⁸U]_AR of dissolved water from exogenously recharged water is usually above 1, increasing with an increase in latitude and altitude (Chabaux et al., 2008,2011; Zhang et al., 2019; Zhao et al., 2020). In such lake environments, lacustrine sedimentary secondary carbonate minerals and interannual biogenic carbonate shells are expected to provide data valuable for reconstructing lake chronologies.