The carbonate δ¹⁸O_carb values in lake sediments are generally controlled by the δ¹⁸O of lake water, water temperature, and disequilibrium effects (Bowen, 1990; Gasse et al., 1991; Yu and Eicher, 1998; Leng and Marshall, 2004). Tian et al. (2003) analyzed the δ¹⁸O of modern precipitation water derived from the Delingha, where it is nearby the Hurleg Lake; the results revealed a strong temperature influence on precipitation isotopic composition in the Qaidam Basin, with higher temperature corresponding to higher isotopic value in precipitation. However, water temperature affects the δ¹⁸O fractionation from lake water into carbonate, which could induce an equilibrium fractionation coefficient of −0.25‰ per °C during calcification (Epstein et al., 1953). The changes in δ¹⁸O values of the HL18-3 core sediments reached 6.87‰, which led to an approximately 27.5°C temperature difference, and there is no evidence for such a high-amplitude change. Therefore, the variabilities of δ¹⁸O_carb cannot account for the variation in the water temperature.

The disequilibrium effects were proven to be negligible at Gengahai Lake on the northeastern Tibetan Plateau (Qiang et al., 2017) and were rarely discussed when evaluating the paleoclimate. Therefore, the δ¹⁸O_carb records of the Hurleg Lake could have been controlled primarily by variations in the δ¹⁸O of lake water (δ¹⁸O_Lw). There is another fact that largely parallel first-order variations in the δ¹⁸O values derived from various carbonate species suggest that all the δ¹⁸O records provide a reasonable approximation of lake water δ¹⁸O (Apolinarska and Hammarlund, 2009; Clegg and Hu, 2010; Qiang et al., 2017). In Hurleg Lake, water samples have an average δ¹⁸O value of −4.9‰ (n = 15) (He et al., 2016); besides, water isotope measurements of 13 samples taken from Hurleg Lake between 2004 and 2007 have a range from −6.6% to −3.3‰ for δ¹⁸O (Zhao et al., 2010). The average value of δ¹⁸O_carb in the HL18-3 core is −5.16‰ (n = 65) (Table 3), and the average value of δ¹⁸O_carb from the upper 50 cm of the HL18-3 core is −6.20‰ (n = 6); the δ¹⁸O_carb values are similar to the δ¹⁸O in the lake water. Therefore, it is considered that the changes in δ¹⁸O_carb can represent the variation of δ¹⁸O in lake water. Hence, factors contributing to δ¹⁸O_Lw will in turn affect the composition of δ¹⁸O_carb. Among these factors, the δ¹⁸O values of the regional rainfall and lake hydrology are the most significant ones (Zhang et al., 2011). Meteorological data from the Delingha weather station show that the lake area of Hurleg Lake was significantly correlated with the annual runoff of the Bayin River (R = 0.78), while the latter is closely related to the mean annual precipitation (Fu et al., 2008; Fan et al., 2014). On the other hand, compared to the annual runoff of the Bayin River, meteorological analyses show that melting ice and snow from glaciers are in low proportion (Fu et al., 2008; Fan et al., 2014). Therefore, the lake area and lake level of Hurleg Lake are dominated by the annual precipitation of the lake catchment (Fan et al., 2014). Correspondingly, the variations in moisture sources and precipitation amounts were considered as significant controls on δ¹⁸O_carb.

In general, moisture transported by the Westerly wind to the region is much more ¹⁸O-depleted than that from the summer monsoon (Araguás-Araguás et al., 1998; Tian et al., 2007; Zhang et al., 2011). Hurleg Lake is located in the transition zone of Westerly wind and Asian monsoon in the past (Winkler and Wang, 1993; Zhao et al., 2011); the variation of moisture source has excellent potential for affecting δ¹⁸O in precipitation. Besides, there is a negative correlation arising from the “amount effect,” whereby the heaviest rainfall is most depleted in δ¹⁸O (Dansgaard, 1964; Zhang et al., 2011). In the monsoon region, significant correlations between δ¹⁸O of precipitation and rainfall amount were observed from the meteorological station (Johnson and Ingram, 2004; Dayem et al., 2010). Although the precipitation amount in the Hurleg Lake region now is low, in the past, the precipitation amounts probably are in large value. This is due to the elevation of three palaeoshorelines in the Holocene which are ∼4, ∼7, and ∼9 m above the present lake level (Fan et al., 2014), and there was a mega-paleolake during the early Pleistocene to late Pleistocene (∼2,800–2,700 m a.l.s) in the north-eastern Qaidam Basin (Huang and Cai, 1987). Thereby, the rainfall amounts could also contribute to the variabilities of δ¹⁸O in precipitation, in turn affecting δ¹⁸O_carb in lake sediments.

In dryland regions, for closed lakes, the water loss is dominated by the evaporation; δ¹⁸O_Lw is therefore strongly influenced by the precipitation/evaporation (P/E) ratio. In previous studies, variations in the δ¹⁸O_carb of lake sediments derived from arid to semiarid regions have always been interpreted as reflecting the balance between regional precipitation and evaporation (P/E) (Lister et al., 1991; Wei and Gasse, 1999; Liu et al., 2007; Zhang et al., 2011). In south Tibet, Tian et al. (2008) suggested that the δ¹⁸O_Lw values in several closed lakes are very sensitive to relative humidity. However, the Hurleg Lake is an open lake, the rapid discharge of a lake system always reduces the lake water residence time and thus decreases the effect of evaporation on δ¹⁸O_Lw (Develle et al., 2010). The Donggi Cona Lake in the Tibetan Plateau provided good evidence, due to the δ¹⁸O_carb values becoming more negative after ∼4.0 cal kyr BP, when the lake became hydrologically open (Mischke and Zhang., 2010). Nevertheless, a recent study on isotopes of Hurleg Lake water displayed that the δ¹⁸O values suffered from a cumulative evaporation effect in which it became stronger from source water to hydrologically open lakes (e.g., Hurleg Lake), then to closed lake systems (He et al., 2016). Further, the results of ²H and ¹⁸O derived from Hurleg Lake water samples fall on a local evaporation line, indicating that the lake water isotopes were influenced by evaporation (He et al., 2016). Therefore, it seems that the impact of evaporation on open lakes cannot be ignored.

Overall, the shifts of δ¹⁸O_carb values from negative to more positive in the Hurleg Lake region over the episodes since the Lateglacial can be interpreted primarily as a decrease in the amounts of precipitation coupled with an increase in the regional P/E ratios, indicating the decrease in regional effective precipitation (Zhang et al., 2011) and representing a dryer climate. Besides, the changes in the moisture source between westerly wind and Asian monsoon may also have played a dramatic role in controlling the δ¹⁸O_carb values.

Organic matter in the lake sediments originating from terrestrial higher plants contains a large number of nitrogen-free biomacromolecules, such as lignin, cellulose, and hemicellulose, which are characterized by carbon enrichment. In contrast, aquatic plants, such as phytoplankton, are characterized by nitrogen enrichment because they are rich in proteins and other substances (Meyers, 1994b; Meyers, 1997; Hedges, et al., 1997). Therefore, the C/N ratio can be utilized to distinguish organic matter derived from various sources, where the C/N ratio ranges of 4–10, 10–20, and >20 correspond to aquatic plants or algae, submerged/phytoplankton or mixed sources of terrestrial and emergent plants, and terrestrial higher plants, respectively (Meyers, 1994b; Jia and Peng, 2003; Hu et al., 2009).

Atmospheric CO₂ exists in three isotopic forms (¹²CO₂, ¹³CO₂, and ¹⁴CO₂); owing to the different activities of light and heavy isotopes in thermal gradients or biochemical reactions, carbon isotope fractionation occurs in plants during transpiration and biosynthesis. The mean δ¹³C_TOC value for C₃ plants is −27‰ (range from −33‰ to −22‰), and for C₄ it is −13‰ (range, −16–9‰) (Cerling et al., 1997). For individual leaf wax n-alkanes, δ¹³C values are ∼6‰–8‰ lower than for bulk tissues (Rieley et al., 1993; Huang et al., 2001); thus, the variabilities of individual leaf wax n-alkanes can reflect changes in C₃ and C₄ plants. C₄ plants prefer to grow in areas with sufficient sunshine, higher temperatures, and drier conditions, primarily composed of warm and dry grassland vegetation (Edwards and Smith, 2010; Strömberg, 2011). C₄ plants have a higher water-use efficiency than C₃ plants and enriched in δ¹³C₃₁ (Edwards and Smith, 2010; Strömberg, 2011). Several recent studies suggested that submerged aquatic plants with enriched δ¹³C may contribute to the concentrations of long-chain n-alkanes (nC₂₇, nC₂₉), while nC₃₁ is a reliable proxy for estimating the variabilities of C₄ plants (Liu and Yang, 2015; Liu and Liu, 2016). Thus, δ¹³C₃₁ was used to reconstruct the evolutionary history of C₄ vegetation (Ma et al., submitted for publication). Generally, higher δ¹³C₃₁ suggests an expansion of C₄ plants, corresponding to a warmer climate.

The generally enriched δ¹⁸O_carb value of fine-grained carbonates suggested a decreased effective precipitation and thus a dryer climate during this period (Figure 5A). However, during this period, the fluctuating upward trend of mean annual precipitation in the Ximencuo Lake northeastern TP coincided with a wetter climatic condition (Figure 7E) (Herzschuh et al., 2014), and the depleted δ¹⁸O of stalagmite calcite from Dongge Cave indicates a strong intensity of Asian monsoon (Dykoski et al., 2005). In addition, the increased C/N ratio revealed that the contribution of terrestrial higher plants to sediments increased (Figure 5B), and the C₄ plants expanded as indicated by more positive δ¹³C₃₁ values (Figure 5C). These results indicated a flourishing of terrestrial and aquatic plants, corresponding to a warm-wet climatic environment. Typically, moisture transported by the summer monsoon to the Tibetan Plateau is more δ¹⁸O-enriched than that of the westerly wind, because of the higher condensation temperature and/or weaker “rain-out” after a shorter air-mass transport distance (Araguás-Araguás et al., 1998; Tian et al., 2007; Qiang et al., 2017). In addition, the δ¹⁸O values of fresh snow samples derived from the Qaidam Basin in January 2000 ranged from −29.4‰ to 20.8‰ (average of −26.2‰), which is significantly more negative than that of Hurleg Lake water (Qiang, 2002; Zhao et al., 2010). Furthermore, a previous study on the northeastern Tibetan Plateau revealed that the hydroclimate was controlled by monsoon-derived, westerly derived, and/or locally recycled moisture in several durations of the Lateglacial (Qiang et al., 2017). Henderson et al. (2010) attributed the depleted δ¹⁸O_carb during the Little Ice Age at Lake Qinghai to the enhanced input of westerly wind. Considering the rapid discharge of the Hurleg Lake system, the δ¹⁸O of lake water was significantly correlated with the annual runoff of Bayin River (R = 0.78), (Fu et al., 2008; Fan et al., 2014). The dramatic hydro-variation of the δ¹⁸O signal induced by the transformation of westerly wind and monsoon would be recorded by the δ¹⁸O_carb at Hurleg Lake. Therefore, the positive shift in δ¹⁸O_carb during 16.1–14.8 cal kyr BP in the Hurleg Lake may be dominated by the strengthened monsoon, and after ∼14.8 cal kyr BP, δ¹⁸O_carb represents multiple changes in monsoon intensity (Figure 5A). The δ¹⁸O_carb derived from the lake sediments in the adjacent areas of Geanggahai Lake in the Qaidam Basin, northeastern Tibetan Plateau, displayed a similar tendency with that of Hurleg Lake (Figure 7B). Combined with other evidence derived from Geanggahai Lake, the data suggest that the Asian summer monsoon may start to dominate the study region at ∼15.3 cal kyr BP (Qiang et al., 2017).

Accordingly, the depleted δ¹⁸O_carb indicated a wet climate from ∼14.8 to ∼12.0 cal kyr BP and enriched δ¹⁸O_carb values during 12.0–11.1 cal kyr BP (Figure 5A), indicating a switch to a drier climate. These findings are in good agreement with the climate reconstructions interpreted by C/N and δ¹³C₃₁ (Figure 5B). The wet-warm conditions at 14.8–12.0 cal kyr BP were likely a response to the wet-warm Bølling/Allerød (B/A) event in the northeast Atlantic Ocean (Bard et al., 2000). At that time, pollen records from Lake Ximencuo on the northeastern Tibetan Plateau inferred that a Cyperaceae-rich high alpine meadow replaced the dry glacial flora at ∼15.5 cal kyr BP, which was then further replaced by alpine meadow and shrubland at 14.0–12.5 cal kyr BP, indicating a humid climate (Herzschuh et al., 2014). The corresponding pollen-based mean annual temperature and precipitation showed a gradual increasing trend in 15.5–12.5 cal kyr BP and showed lower values during 12.5–10.8 cal kyr BP (Figures 7D,E) (Herzschuh et al., 2014). In addition, the U^K ₃₇ record at Lake Qinghai inferred a significantly increase in temperature since 15.1 cal kyr BP, which remained at relatively high temperature until 12.6 cal kyr BP. This was followed by a lower temperature during 12.6–11.7 cal kyr BP (Figure 7C) (Hou et al., 2016).

After that, the 12.0–11.1-cal kyr BP dry interval at the Hurleg Lake may be linked to the Younger Dryas period (Shakun and Carlson, 2010), which markedly affected most regions of the Northern Hemisphere, including the Tibetan Plateau. Lake Dongxiong Co. in the southwestern Tibetan Plateau experienced a 3.8°C drop in temperature during 12.3–11.4 cal kyr BP (Ling et al., 2017). Other lake records on the Tibetan Plateau, such as Qinghai Lake (Ji et al., 2005), Lake Sumxi Co. (Gasse et al., 1991), and Lake Nam Co. (Günther et al., 2015), revealed a commonly cold-dry/wet climate lasting from ∼12.9 to ∼10.0 cal kyr BP.

Soon after 11.1 cal kyr BP, the δ¹⁸O_carb values rapidly decreased and reached a relatively low value at ∼10 cal kyr BP and then increased (Figure 5A), inferring a wet climate in the early Holocene, which may be induced by the enhanced Asian monsoon (Figure 8E) (Dykoski et al., 2005). Since ∼8.3 cal kyr BP, δ¹⁸O_carb quickly depleted and then gradually enriched, reflecting a wetter climate than the previous stage, and a drier climate in the late mid-Holocene (Figure 6A). During this interval, the δ¹³C₃₁ values were generally high in 11.1–9.2 cal kyr BP and then decreased to low values during 9.2–8.3 cal kyr BP. This corresponds to a warm climate at the beginning of the Holocene, which was interrupted by a cold interval around 8.3 cal kyr BP (Figure 5C). The ∼8.3-cal kyr BP cold event has been commonly acknowledged in large numbers of studies on the Tibetan Plateau (Herzschuh et al., 2006, 2014; Mischke and Zhang, 2010; Zhang and Mischke, 2009) and in the Hurleg Lake (Zhao et al., 2007; Zhao et al., 2010; Zhao et al., 2013; He et al., 2016). The precipitated carbonate δ¹⁸O of the HL05-2 core showed a similar variation trend to the δ¹⁸O_carb in the HL18-3 core during 9.2–8.3 cal kyr BP (Figure 6B), which signified a dry climate (Zhao et al., 2010). Moreover, the lake level reflected by the carbonate content suggested that of the lake contracted notably at ∼7.9 cal kyr BP (Figure 6C) (Zhao et al., 2010). The reconstructed salinity via alkenone-based %C_37:4 exhibited relatively high values at 7.7–8.2 cal kyr BP (Figure 6E), indicating a dry climate (He et al., 2016). Besides, U^k' ₃₇-derived ΔT fluctuated rapidly at low values during this period, which then increased distinctly after ∼8.0 cal kyr BP (Figure 6D) (Zhao et al., 2013). Furthermore, pollen concentrations were low during intervals (Figure 6F) (Zhao et al., 2007; Zhao et al., 2013), and the C₄ plant indicator was significantly reduced (Figure 5C, Ma et al., submitted for publication); correspondingly, the organic matter percentage declined during 8.3–8.7 cal kyr BP (Figure 6G) (Zhao et al., 2010). When considering chronological errors, we determined the occurrence of a vital cold-dry event, which may be related to the frequently recorded 8.2-cal kyr BP cold event in the North Atlantic and many parts of the world (Johnsen et al., 1992; Alley et al., 1997; Thomas et al., 2007; Hede et al., 2010).

The enriched δ¹³C₃₁ value after ∼8.3 cal kyr BP indicates a recovery of C₄ plants (Figure 5C). From ∼8.0 to ∼6.2 cal kyr BP, the low δ¹⁸O_carb value suggested a generally wet climate, corresponding to a high lake level, reported by Fan et al. (2014), The excessively humid climate did not favor C₄ plant growth; thus, the δ¹³C₃₁ value was depleted, representing a decrease in C₄ plant biomass (Figures 5A,C). Afterward, the higher δ¹⁸O_carb values during 6.2–4.8 cal kyr BP suggest a gradual drying climate, which is beneficial to the growth and expansion of C₄ plants, as reflected by the enriched δ¹³C₃₁ value (Figures 5A,C). The changes in precipitated carbonate δ¹⁸O of the HL05-2 core (Figure 6B) (Zhao et al., 2010) and Genggahai Lake (Figure 7A) (Qiang et al., 2017) were consistent with the δ¹⁸O_carb of the Hl18-3core during this period (Figure 6B) (Zhao et al., 2010).

In the late Holocene, the δ¹⁸O_carb value showed a general depleted trend, similar to that of other climate indices, such as C/N and δ¹³C₃₁ (Figures 5A–C). The δ¹⁸O_carb value was enriched during 4.8–4.3 cal kyr BP, while δ¹³C₃₁ decreased rapidly, indicating a contraction of C₄ plants. All of these proxies signified a strong abrupt cold-dry event (Figures 5A,C). The temperature indicator U^k' ₃₇-derived ΔT showed low values during 4.9–4.3 cal kyr BP. Moreover, the carbonate content declined revealing a low lake level, and a higher salinity appeared at ∼4.6 cal kyr BP, suggesting a dry climate (Figures 6C–E) (Zhao et al., 2013; Zhao et al., 2010; He et al., 2016). The pollen concentration and organic matter percentage were in low value from ∼4.2 to 3.7 cal kyr BP (Figures 6F,G) (Zhao et al., 2007; Zhao et al., 2013), which demonstrated a dry and deteriorated climate. These results generally signify a cold-dry event, correlated with a 4.2-cal kyr BP event (Bond et al., 1997; Mischke and Zhang, 2010; Hou et al., 2016). The positive δ¹⁸O_carb values observed in the Genggahai Lake during ∼5.0–∼3.5 cal kyr BP, as well as the decreased precipitation in Ximencuo Lake on northeastern Tibetan Plateau during 4.9–3.8 cal kyr BP, indicated that a hydroclimate was dry, while the temperature in the adjacent area of Hurleg Lake, such as the Qinghai and Ximencuo lakes, decreased dramatically (Figures 6C,D) (Herzschuh et al., 2014; Hou et al., 2016). In addition, the sea ice coverage in the Barents Sea increased significantly (Figure 7F) (De Vernal et al., 2013), sea surface temperature decreased (Lea et al., 2003), and Greenland ice core δ¹⁸O (Stuiver et al., 1995) notably decreased during this period, indicating a global cold event at ∼4.2 cal kyr BP.

After the rapid cold-dry event, the climate conditions improved slightly around 4.2–3.0 cal kyr BP, as shown by a negative δ¹⁸O_carb (Figure 5A), indicating a higher input of effective precipitation. However, the C/N ratio revealed a higher input of terrestrial plants, while the depleted δ¹³C₃₁ pointed to less C₄ plants (Figure 5C). Here we argued that although the C₄ plants decreased, with the more rapid expansion in C₃ plants, it led to a high value of the C/N ratio. In the meantime, the decrease of C₄ plants also suggests that the climate is gradually turning cold (Figure 5C). Subsequently, the climate displayed a generally cold-wet trend, as revealed by δ¹⁸O_carb, C/N, and δ¹³C₃₁ (Figures 5A–C). Fan et al. (2014) reported that a high lake level occurred during 2.2–1.4 cal kyr, and Song et al. (2020) also revealed a moderately wet climate after 2.5 cal kyr BP by sediment grain size and δ¹³C of authigenic carbonate in the Hurleg Lake.