Factors controlling the oxygen isotopic composition of lacustrine authigenic carbonates in Western China: implications for paleoclimate reconstructions

In the carbonate-water system, at equilibrium, the oxygen isotopic composition of carbonate is dependent not only on the temperature but also on the isotopic composition of host water in which the carbonate is formed. In this study, lake surface sediment and water samples were collected from 33 terminal lakes in Western China to evaluate controls on the oxygen isotopic composition of lacustrine authigenic carbonates (δ18Ocarb) and its spatial distribution. Our results show that water oxygen isotopic composition (δ18Owater) rather than lake summer water temperature (Twater), is the main determinant of δ18Ocarb, irrespective of whether oxygen isotope equilibrium is achieved. There are significant linear correlations between δ18Ocarb and elevation, as well as that between δ18Ocarb and latitude for lakes located on the Tibetan Plateau. In Western China, the spatial distribution of δ18Ocarb is consistent with that of δ18Owater, and is ultimately controlled by the isotopic composition of local precipitation (δ18Oprecipitation) that depends on the source of water vapor. Therefore, changes in δ18Ocarb can be predominantly interpreted as variations of δ18Owater, which in turn represent changes in δ18Oprecipitation for paleoclimate reconstructions in this region, and may be relevant to studies of other areas.

The carbonate oxygen isotope geothermometer has been one of the most important tools for reconstructing paleotemperatures since it was established by Urey 1 and McCrea 2 . In systems that are in thermodynamic equilibrium, an increase in temperature with increased atomic vibration frequency leads to decrease in isotopic fractionation between the reactant water and the carbonate mineral that precipitates from it [3][4][5] . Accordingly, a number of empirical relationships between carbonate oxygen isotopic composition and temperature have been established [5][6][7][8] , and are widely applied to lakes in paleoclimate studies to reconstruct water temperature [9][10][11] .
Our work on oxygen isotope systematics in modern lake carbonates is motivated by three issues. First, the fundamental precondition to applying an isotopic geothermometer is that carbonate precipitating from water is in oxygen isotope equilibrium 1,12 . The assumption of equilibrium has not been tested in detail in modern lacustrine authigenic carbonates and there are discrepancies relating to equilibrium fractionation factors and kinetic fractionation processes among theoretical calculations, experimental measurements, and modern natural observations 13,14 . In slowly precipitating experiments from Kim and O'Neil 7 , the authors judged the smallest fractionation factors to be the best representations of equilibrium fractionation factors, despite the fact that fractionation factors increased with increasing initial concentrations of metal ion and bicarbonate at a certain

Results
Summer water surface temperature calculation. Logged Mean Summer Water Temperature (T LMSW ) from June to August was calculated using the data recorded by the on-site data loggers retrieved from twelve lakes (Supplementary Table S1 online). We derived a regression based on the relationship between Midday Temporal Water Temperature (T MTW ), measured manually using a mercurial thermometer in the field, and the T LMSW from the loggers that were able to be retrieved from the lakes: The T LMSW is positively correlated to the T MTW (Fig. 2). For 21 sites where water temperature loggers were lost in the field, we determined the Calculated Mean Summer Water Temperature (T CMSW ) by applying Eq. (1) to the T MTW values for the lakes without loggers. Thus, in this study, lake summer water surface temperature (T water ) is either T LMSW (lakes with loggers) or T CMSW (lakes without loggers). T water ranged from 9.8 to 25.6 °C (Supplementary Table S1 online). Lake surface water and surface sediment information. The salinity of the studied thirty-three lake surface waters ranged from 354.18 to 87,991.23 mg/L. The pH values ranged from 7.89 to 9.81. The saturation index (SI) value of calcite, aragonite, or dolomite exceeds 0 for each sample (Supplementary Table S2 online). Although there is no relevant data to calculate the SI value of monohydrocalcite, we assumed the waters were supersaturated in monohydrocalcite in the cases where rapid deposition occurred. The oxygen and hydrogen www.nature.com/scientificreports/ isotopic composition of lake surface water (δ 18 O water and δD water ) ranged from − 8.82 to 5.65‰ (VSMOW) and from − 79.12 to 11.22‰ (VSMOW), respectively (Supplementary Table S2 online). In absence of detrital and biogenic carbonate, the fine sieved carbonate smaller than 45 μm can be characterized as authigenic carbonate, which is chemically precipitated in lake water 37 . X-ray powder diffraction (XRD) analyses show that there are nine pure calcite samples and twenty-four mixed mineralogy samples in this study (Supplementary Table S3 and Fig. S1 online). The oxygen isotopic composition of lacustrine authigenic carbonate (δ 18 O carb ) was calculated based on a stable isotope mixing model. δ 18 O carb spanned a relatively large range  Table S1 online. The software BIGEMAP (https ://www.bigem ap.com) was used in this study to download the satellite imagery from Google Earth (Map data: Google, Maxar Technologies). The spatial location, the data layer creation and "shape" vector format file generation were performed in software ArcGIS 10.2.  Relationships between 1000lnα (carb−water) , t water , δ 18 o carb and δ 18 o water . As shown in Fig. 3a, there is no statistically significant correlation between 1000lnα (carb−water) and T water : All points are scattered on both sides of the fitted line, where 18 points distribute out of the 95% confidence interval. Thus, an initial inspection indicates the 1000lnα (carb−water) values cannot be used to infer lake summer water temperature.
(2) 1000lnα (carb−water) = − 0.12 ± 0.10T water + 30.92 ± 1.74(n = 33, r = 0.21, P = 0.2457) 18 O water − 2.76 ± 0.45(n = 33, r = 0.78, P < 0.0001) Figure 3. Comparison of (a) T water and 1000lnα (carb−water) ; (b) δ 18 O water and δ 18 O carb values of the thirty-three lakes in this study. Black points refer to water temperatures directly recorded by on-site water temperature loggers (T LMSW ). Blue points refer to water temperatures for sites without data loggers and are calculated using the Eq. (1) and are reported as Calculated Mean Summer Water Temperature (T CMSW ). Similar results are obtained using both types of temperature data. Solid lines are least-square linear regression lines. Error bars show standard deviations in each sample. Dotted lines show 95% confidence intervals.    . Relationship between oxygen isotopic composition of carbonate and host waters in (a) a combination of isotope data collected from published papers and thirty-three lakes, and in (b) lake surface sediments. Black crosses are data for thirty-three lake surface water samples analyzed in this study. Also shown are data for laboratory synthetic carbonates from experiments, lake surface sediment samples, and other field-collected samples compiled from published papers (Supplementary Table S4

Discussion
Two critical factors controlling the oxygen isotopic composition of carbonate mineral are the temperature of carbonate formation and the oxygen isotopic composition of the solution from which carbonate minerals precipitate 5 . In this section, we discuss the effects of these two factors respectively and spatial distributions of the oxygen isotopic composition of lacustrine carbonates in Western China. In this study, the carbonate samples provide integrated climate signals cover several years as the sedimentation rate of surface sediment ranges from 0.01 to 0.3 cm/yr throughout our sampling locations [37][38][39][40][41] . In general, lacustrine authigenic carbonates precipitate in summer, when the carbonate saturation of lake water peaks and carbonate solubility is simultaneously depressed in the epilimnion 5,9,42,43 . Variations in water temperatures averaged over several summers in recent years usually less than ± 1 ~ 2 °C at a certain lake, especially for the lakes located on the Tibetan Plateau 44,45 . Therefore, the measured summer water temperatures could represent the temperatures when authigenic carbonate samples were precipitated. In large closed lake systems, variations in water isotope composition caused by precipitation or evaporation are usually homogenized by buffering of large lake volume 5,24 . Therefore, a large lake with long water residence time could 'average out' short-term changes in isotope compositions and instead reflects relatively long-term isotope compositions under similar climate and hydrological conditions 5,46,47 . As the size of sampled thirty-three lakes is relatively large, and the relative humidity in Western China has not changed greatly in recent years 48,49 , variations in water isotope values during the course of one or several summers may not significant 50 . Because we do not have longitudinal data on lake water isotope values at our lakes, we assume that the measured isotope values of water samples that collected at the lake center and at the same time with sediment samples could be considered as long-term average compositions during the summer when authigenic carbonates were precipitated.
For equilibrium carbonate precipitation, oxygen isotope fractionation is directly controlled by thermodynamics, and the isotope fractionation factor is a function of temperature 1, 2 . However, there is no statistically significant correlation between the 1000lnα (carb−water) and water temperatures in natural lake settings in this study (Fig. 3a). Since calcites from the Devils Hole and Laghetto Basso are considered to be most representative of true thermodynamic equilibrium, we further compared our results with the equilibrium baseline defined by Daëron et al. 17 and other published temperature dependent oxygen isotope equilibrium fractionations [5][6][7] . We modeled water temperatures (T CW ) by applying published temperature calibrations to the 1000lnα (carb−water) values of our lakes, and compared modeled T CW with independently measured T water . As shown in Fig. 6a, the modeled T CW ranges from 3.8 to 68.2 °C, which is far beyond T water (9.8-25.6 °C) of our lakes. The slope and intercept of 1000lnα (carb−water) -T CW regression lines are lower than that of 1000lnα (carb−water) -T water regression line (Fig. 6a). The discrepancy between our results and published calibrations is salient as shown in Fig. 6b, in which most of modeled T CW using the Daëron et al. 17 calibration are higher than the T water and no statistically significant correlation existed between T CW and T water (P = 0.2199). This suggests that factors other than temperature are contributing to oxygen isotope fractionation observed in lacustrine authigenic carbonates in this study. The discrepancies between our results and previous studies may be attributed to carbonate disequilibrium precipitation.
Lacustrine authigenic carbonates form in a mixture of dissolved inorganic carbon (DIC) species in lake water. At equilibrium, the oxygen isotopic composition of precipitated carbonate closely approximates that of DIC at a certain temperature 18,51 . As lake water is the largest reservoir of oxygen isotopes for DIC species and precipitated carbonates, we employed a mass balance calculation 13,52 to calculate the expected oxygen isotope fractionation factor between DIC and water (1000lnα (DIC−H2O) ) under equilibrium fractionation, in order to investigate whether equilibrium oxygen isotope fractionation are achieved in 33 lakes in Western China. For the calculation of expected equilibrium (1000lnα (DIC−H2O) ) values, we neglect CO 2(aq) because there should not be a substantial influence of CO 2 degassing on the DIC pool in natural lake settings 53 . The mass balance equation is: where X denotes the molar fractionation of the DIC species which is determined by PHREEQC v. 2.18.00. software program 54 , and 1000lnαi denotes the individual fractionation factor between the DIC species and lake water reported by Beck et al. 51 at a certain summer water temperature in each lake.
We compared our experimental 1000lnα (carb−water) values to expected equilibrium (1000lnα (DIC−H2O) ) values ( Supplementary Fig. S2 online). There is no statistical linear correlation between the experimental 1000lnα (carb−water) values and the expected equilibrium (1000lnα (DIC−H2O) ) values (r = 0.23, P = 0.2033). Most of the 1000lnα (carb−water) values are lower than the expected equilibrium (1000lnα (DIC−H2O) ) values (Supplementary  Table S3 online). The offsets ranging from 0.19 to 9.25 indicate that the oxygen isotope values of precipitated lacustrine authigenic carbonates may not in equilibrium with lake waters in this study, and the temperature is not the primary control on carbonate δ 18 O values in natural lake settings.
Carbonate disequilibrium precipitation may be influenced by multiple mixed kinetic fractionation processes that originate from kinetic fractionations during the exchange of oxygen isotopes between water and DIC species, or between DIC species and carbonate, or a combination of both of these factors 18,55 . With respect to lakes, there are two main factors that can lead to the disequilibrium fractionation that we observe; these may be complicated by multiple environmental controls and processes at a given site.
First, the pH value of lake water is an important factor in carbonate isotope fractionation processes. The pH value of the solution determines the concentration of each DIC species, which in turn controls the relative proportions of DIC species participating in carbonate growth at a certain temperature 15,56,57 . The oxygen isotope fractionation between water and DIC, as well as that between DIC and precipitated carbonates, will decrease with increasing pH, as the dominant DIC species changes from CO 2(aq) to CO 2− www.nature.com/scientificreports/ Fig. S3 online). Although several of the data points can be explained by a combination of temperature and pH, however, more than half of the samples fall out of the range predicted by the model (Supplementary Fig. S3 online), indicating pH values of lake water are not the dominant factor controlling the oxygen isotopic composition of lacustrine authigenic carbonates at many of the sites in Western China. Second, the oxygen isotope exchange between DIC species and water is a rate-limiting step for equilibrium 13,58 . High carbonate growth rates may result in kinetic fractionation of a different magnitude 21,56,59 . The slower the carbonate is formed, the more likely the isotope fractionation between water and carbonate is to be close to equilibrium. Authigenic carbonates are thought to form in lacustrine settings relatively rapidly 37,60 , especially for carbonates precipitated from saturated solutions (SI > 0) as in this study. Since the growth rates of carbonate samples in this study are, as with many carbonates, likely to be higher than that of slowly precipitated calcites collected from the Devils Hole and Laghetto Basso, the time for oxygen exchange between the DIC and water might be insufficient to attain the equilibrium fractionation 16,17 . In this case, the DIC species with 16 O isotopologues tend to preferentially participate in oxygen isotopic exchange, leading to the oxygen isotope composition of formed carbonates being lighter than theoretical values 20,55 . Therefore, kinetic fractionations caused by pH Figure 6. Relationship between (a) lacustrine authigenic carbonate oxygen isotope fractionation factors (1000lnα (carb−water) ) and lake summer water surface temperature (T water ) and modeled water temperatures (T CW ) by applying several published temperature calibrations [5][6][7]17 ; and (b) lake summer water surface temperature (T water ) and modeled water temperatures (T CW ) using the Daëron et al. 17  www.nature.com/scientificreports/ effect and growth rate effect might lead to large deviations of fractionation factors between expected equilibrium and measured results in natural lake settings, resulting in the lack of a significant correlation between 1000lnα (carb−water) and T water in this study. In order to compare the relationship between δ 18 O carb and δ 18 O water , we compiled more than five hundred oxygen isotope values of different types of carbonates from published papers ( Fig. 4 and Supplementary Table S4 online). Although some of the published isotope data were assumed to have reached oxygen isotopic equilibrium, for most of the data, it is unclear whether oxygen isotopic equilibrium was reached when carbonate precipitated from host water. As shown in Fig. 4, the linear correlation between δ 18 O carb and δ 18 O water values is significant both in all carbonate data (n = 541, r = 0.91, P < 0.0001) and in lake surface sediments (n = 121, r = 0.95, P < 0.0001). These results indicate that higher δ 18 O water values correspond to higher δ 18 O carb values, even though hydrologic conditions differ between records and may be complex, and oxygen isotope equilibrium is not necessarily attained during carbonate precipitation. The positive correlation between δ 18 O water and δ 18  As for the oxygen isotopic composition of lake water, it mainly depends on the changes of the isotopic composition of precipitation and local evaporation, in closed lake basins where rivers in the catchment are supplied by precipitation and no surface or groundwater output exists 61 . For terminal lakes, covariant trends between carbonate δ 18 O and δ 13 C reflect isotope enrichment caused by kinetic fractionation during evaporation 9,62,63 . As shown in Fig. 7a, there is no significant linear correlation (P = 0.4108) between δ 18 O carb and δ 13 C carb for the lakes in this study. But if we classify samples by the location of lakes, the correlations between δ 18 O carb and δ 13 C carb become more considerable. Linear correlation coefficients (r) range between 0.35 and 0.92 depending on the location of the lake (Fig. 7b). We infer that the influence of evaporation on δ 18 O carb may be less than the influence of precipitation isotopic composition on δ 18 O carb in Western China. This inference is also supported by spatial variations of lake water δD and δ 18 O at the 33 sites sampled for this study 64 . Feng et al. 64 discussed the relationship between isotopic composition of lake water (δ 18 O water and δD water ) and local precipitation (δ 18 O precipitation and δD precipitation ) in Western China, and investigated influences of lake latitude, elevation, and lake water salinity on δ 18 O water and δD water in detail. In their results, isotope enrichment by local evaporation, of differing magnitudes depending on location, is also recorded by the Local Evaporation Line (LEL) that shifts to the right of the Local Meteoritic Water Line (LMWL) (Fig. 7c). However, the spatial distribution of lake water isotopes is in accordance with that of precipitation isotopes, with heavier δ 18 O precipitation and δD precipitation corresponding to more enriched δ 18 O water and δD water at same region in Western China (Fig. 8). Feng et al. 64 concluded that δ 18 O water in these lakes located in Western China is mainly controlled by δ 18 O precipitation that depends on the source of water vapor, while local evaporation, lake elevation and latitude have less influence on δ 18 O precipitation and δ 18 O water .
As changes in δ 18 O water dominate the variations in δ 18 O carb , the spatial distribution of δ 18 O carb could be inherited from that of δ 18 O water in Western China. Based on Rayleigh fractionation, isotope values for precipitation become gradually depleted when water vapor climbs high mountains 65,66 . As shown in Fig. 5a,b, correlation coefficients (r) of elevation are 0.76 (P < 0.0001) and 0.88 (P < 0.0001) for carbonate and lake water, respectively, demonstrating that altitude effect is significant on the Tibetan Plateau. But δ 18 O carb variances caused by the altitude effect is not significant (r = 0.78, P = 0.0243) at the Northwestern Xinjiang and Inner Mongolia regions where lake elevation is below 3000 m. On a worldwide scale, water vapor originates from the tropical ocean 65 . The latitude effect usually results in heavy oxygen isotope depletion of precipitation when water vapor is transported from southern to northern regions 65,67 . Although δ 18 O water and δ 18 O carb are negatively correlated to elevation, they are positively correlated to latitude for lakes located on the Tibetan Plateau. It indicates that the variation of δ 18 O water and δ 18 O carb may not be explained by the latitude effect in Tibetan Plateau. For the southern Tibetan Plateau, precipitation with negative isotope value is mainly originated from the Bay of Bengal and the Arabian Sea 68 . But the oceanic water vapor is usually blocked by the Himalayas and Tanggula Mountains and can hardly arrive northern part of the plateau 49,69 . Lakes located on the northern Tibetan Plateau are under the control of dry continental air masse with enriched heavy isotopes 64 . As a result, δ 18 O water and δ 18 O carb values of lakes located on the northern Tibetan Plateau are higher than these of lakes located on the southern Tibetan Plateau. Since δ 18 O precipitation values provided by the westerlies and East Asia summer monsoon are different 67 , the insignificant Lat/δ 18 O carb coefficient (r = 0.33, P = 0.4238) for lakes located in the Northwestern Xinjiang and Inner Mongolia regions could be also attributed to different vapor sources. As shown in Fig. 5e,f, the correlations between lake longitude and δ 18  In natural lacustrine settings, the temperature dependent oxygen isotope equilibrium fractionation between lacustrine carbonate and lake water can be complicated by multiple environmental controls and processes, such as ionic saturation, DIC speciation, growth rate, and other factors 70,71 . It is likely that the crystallization and precipitation of lacustrine carbonate occurs under non-equilibrium conditions 17 . As such, even though temperature is an essential factor controlling the oxygen isotopic composition of carbonate minerals, it is questionable to reconstruct water temperature using lacustrine authigenic carbonates, without a detailed discussion of whether isotopic equilibrium was attained. In contrast, our results indicate that the isotopic composition of lacustrine authigenic carbonate depends on that of host water, regardless of whether the isotope equilibrium conditions are reached. Therefore, in paleoclimate reconstructions, changes in oxygen isotopic composition of lacustrine authigenic carbonate from terminal lakes in Western China can potentially be interpreted as the oxygen isotopic composition of lake water, which in turn indirectly reflects variations of the oxygen isotopic composition of precipitation, assuming intra-annual changes in temperature were relatively small. Furthermore, the spatial distributions of lacustrine authigenic carbonate oxygen isotope values could reflect different water vapor origins in Western China.
Overall, the main conclusions we reached are: (1) Temperature is associated with a relatively small fraction of the observed variance in δ 18 O carb and 1000lnα (carb−water) in natural lake settings in Western China; (2) Many factors may lead to kinetic oxygen isotope fractionation during authigenic carbonate precipitation in lacustrine settings, including pH and growth rate-related effects. These factors can account for a larger fraction of the variance in δ 18 O carb and 1000lnα (carb−water) in these samples; (3) A positive correlation between δ 18 O water and δ 18 O carb is observed in the 33 lakes located in Western China. δ 18 O water is the dominant factor governing δ 18 O carb , regardless of whether the isotope equilibrium conditions are reached during the precipitation of authigenic carbonates; (4) The spatial distribution of δ 18 O carb is consistent with that of δ 18 O water and δ 18 O precipitation , and is ultimately controlled by water vapor source in Western China; (5) Under either the presence or absence of isotope equilibrium, changes in δ 18 O carb from terminal lakes in Western China can be predominantly interpreted as variations of δ 18 O precipitaion , instead of changes in temperature. This provides an important basis for future paleoclimatic reconstructions using the carbonate oxygen isotope proxy in lacustrine authigenic carbonates.

Methods
Sampling of lake surface sediment and water. In July and August 2016, surface sediment and water samples were collected from thirty-three lakes located in Western China. To ensure that the samples were not influenced by hydrological or human disturbance, samples were collected at the lake center for smaller lakes and at least 2 km away from the shore for larger lakes.
In each lake, the upper most 0.5 cm of surface sediments were collected using a stainless grab and were placed in leak proof plastic bags. At the same site, surface water samples were collected at a depth of ~ 50 cm below the water surface. Water samples were collected and stored in 500 ml high-density polyethylene (HDPE) bottles which were initially washed three times using the lake water. The bottles were completely filled with water samples and sealed with a cap secured with plastic electrical tape to avoid evaporation or any isotopic exchange with air. Sediment and water samples were kept cool in the field and were then stored at 4 °C in Capital Normal University, China.
T MTW was measured once for each lake at around 2 p.m., during the warmest time of day, when the sediment and water samples were collected in the field. T MTW was manually measured using a mercurial thermometer at 50 cm below the water surface in the same location of water sampling. HOBO U22 Water Temperature Pro v2 data loggers were also set at a depth of 50 cm below the water surface of each lake. Temperature data was collected at 15-min intervals over the course of one year. We returned in the following summer and successfully retrieved 12 data loggers, while the rest of the loggers were lost. Figure 7. (a) Relationship between δ 18 O carb and δ 13 C carb in this study. No significant linear correlation between δ 18 O carb and δ 13 C carb (P = 0.4108) is observed for the thirty-three samples. Black points refer to water temperatures directly recorded by on-site water temperature loggers (T LMSW ). Blue points refer to water temperatures for sites without data loggers and are calculated using the Eq. (1) and are reported as Calculated Mean Summer Water Temperature (T CMSW ). (b) Relationship between δ 18 O carb and δ 13 C carb in this study. Samples are classified according to the position in which each lake located. Red points denote lakes located on the Northern Tibetan Plateau, light blue points denote lakes located on the Southern Tibetan Plateau, green points denote lakes located in the Northwestern Xinjiang region, pink points denote lakes locate at the Inner Mongolia region. The solid lines are least-square linear regression lines. (c) Oxygen and hydrogen isotopic composition of thirty-three water samples measured for this study 64 . Triangles are modern local precipitation in the study areas which are derived from the Online Isotopes in Precipitation Calculator 85 . Circles are lake surface water which coded in different colors according to the position in which each lake is located. The pink dashed line is the Global Meteoritic Water Line (GMWL) 86  www.nature.com/scientificreports/ Water chemistry and stable isotope analyses. Lake surface water samples were analyzed for K + , Ca 2+ , Mg 2+ , SO 2− 4 , Cl − , HCO − 3 , and CO 2− 3 concentrations at the Qinghai Institute of Salt Lakes, Chinese Academy of Sciences, China. All analyses for major ions in this study followed procedures of the Qinghai Institute of Salt Lakes 72 . Cl − concentrations were determined by AgNO 3 potentiometric titration, with a precision of ± 0.1%. CO 2− 3 and HCO − 3 concentrations were analyzed by HCl titration, with a precision of ± 0.3%. Concentrations of SO 2− 4 were determined by gravimetric methods through precipitation of BaSO 4 . Concentrations of K + were measured by gravimetric methods through precipitation of potassium tetraphenylborate [KB(C 6 H 5 ) 4 ]. Ca 2+ and Mg 2+ concentrations were measured by ethylene diamine tetraacetic acid (EDTA) titration with errors of ± 0.5%. Na + concentrations were calculated by charge balance: where N represents the ionic equivalent value. The analytical precision for major cations and anions is better than ± 2%. Water salinity was calculated based on the concentrations of major aqueous ions. pH values were measured in the field during sample collection with a Mettler SevenGo2-ELK. At each lake, the probe was calibrated three times using standard pH calibration solutions (4, 6.86 and 9.18 at 25 °C). pH values of standard calibration solutions were also adjusted for measured water temperatures at the field sites. The distribution of species and carbonate SI values were calculated using the equilibrium geochemical speciation/mass transfer model PHREEQC v. 2.18.00. software program 54 with the speciation model wateq. database. δ 18 O water and δD water were conducted at the Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, China using an LGR DLT-100 Liquid Water Isotope Analyzer (Los Gatos Research, Inc., Mountain View, CA, USA). Calibration of the measurements used three internal LGR standards (δ 18 O: − 2.80‰, − 7.69‰, and − 13.10‰; δD: − 9.5‰, − 51.0‰, and − 96.4‰). δ 18 O water and δD water were reported relative to VSMOW. The measurement accuracy was typically better than ± 0.1‰ for δ 18 O water and ± 0.5‰ for δD water .
Sediment sample pretreatments. The wet surface sediment samples were soaked in deionized water for about 2 h and then wet sieved with a 350-mesh (45 μm) sieve. Materials exceeding 45-μm containing detrital mixtures and biogenic carbonates (containing primarily ostracods) were filtered out 37,73 . Only fine sieved fractions smaller than 45 μm were collected, frozen in a refrigerator overnight and then vacuum freeze-dried for 48 h using the Boyikang FD-1A-50 Freeze Dryer at approximately − 50 °C (30 Pa), until the samples were dried. Around 2 g of each sieved sediment was ground using agate mortar and pestle, and stored in a desiccator. Sediment X-ray powder diffraction analyses. Around 0.5 g of sediment powder were loaded into a plastic sample holder and the surface of the powder was smoothed prior to XRD measurements that were performed at the Qinghai Institute of Salt Lakes, Chinese Academy of Science, China, using a Phillips X-pert Pro X-ray diffraction with Cu K α radiation (λ = 1.5406 Å). The diffraction spectral pattern was measured at a scanning rate of 2° min -1 for 2θ ranging from 10° to 80°. Mineral identification and semi-quantitative analyses were estimated from the bulk mineral diffractograms using the reference-intensity ratio (RIR) matrix-flushing method [74][75][76] aided by the use of an automated search-match computer program X'pert HighScore Plus. The uncertainty of this semi-quantitative analysis was approximately ± 5% (1σ). carbonate oxygen and carbon isotope analyses. The fine sieved sediment samples were treated with 3% H 2 O 2 for 4 h to remove any remaining organic material. Resulting samples were collected on a 0.45 μm cellulose nitrate filter membrane and oven-dried at 40 °C. Depending on instrument sensitivity and carbonate content, the amount of sample used for isotope analyses varied between 12 and 95 mg.
The δ 18 O carb and δ 13 C carb were measured with a Thermo Scientific MAT 253 gas source isotope mass spectrometer at the University of California, Los Angeles, USA from 2017 to 2018. Samples were reacted with 105% phosphoric acid (ρ = 1.92 g/mL) for 20 min on a 90 °C online common phosphoric acid bath system to convert to CO 2 gas for analyses. The liberated CO 2 was successively passed through a dry ice/ethanol trap (− 76 °C) and a liquid nitrogen trap (− 196 °C) to remove water and other compounds. After the initial purification step, the CO 2 was passed through silver wool to remove sulfur compounds and then passed through a Porapak Q gas chromatograph column at − 20 °C to remove any additional contaminants before being transferred into bellows of the mass spectrometer for analysis. Data were collected over 9 acquisition cycles to determine δ 13 C and δ 18 O. A high purity pre-calibrated CO 2 tank was used as a reference gas (From 1/19/2017 to 2/21/2018: Source was Air Liquide with δ 18 O = 19.31‰ VSMOW, δ 13 C = − 3.38‰ VPDB; after 2/21/2018: Source was Oztech with δ 18 O = − 15.84 ‰ VPDB, δ 13 C = − 3.64‰ VPDB), whose composition has been determined by Oztech through comparison with NBS standard gases and CO 2 evolved by acid digestion from NBS-19 and NBS-18. At least three replicates per sample were performed.
For calcite, 18 O/ 16 O fractionation by phosphoric acid digestion at 90 °C was corrected using an acid fractionation factor of 1.00795 77 . For aragonite, an acid fractionation factor value of 1.00854 was calculated by extrapolating the relationship reported by Kim et al. 78 . For dolomite, we used an acid fractionation factor of 1.0093 79 . For monohydrocalcite, we used the same fractionation factor as calcite 80 . For samples that are a mixture of two or three carbonate minerals as determined by XRD, weighted acid fractionation factors were calculated. δ 18 O carb and δ 13 C carb are reported on the VPDB scale. We ran NBS-19 standards. The average measured value for NBS-19 is: δ 13 C = − 2.143 ± 0.021‰ VPDB, δ 18 O = 1.938 ± 0.012‰ VPDB (n = 6). We also ran ETH-1 to 4 as standards 81 . Carbonate standards were analyzed between every 2-3 samples and were prepared and analyzed in the same manner as samples.
Based on the stable isotope mixing model: www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.