Lake Cuoqia is located in Hengduan Mountains in the southeastern Tibetan Plateau, ~20 km southwest of Shangri La County, Diqing Autonomous Prefecture, Yunnan Province (27°24′18.72″ N, 99°46′19.87″ E; elevation: 3960 m; Figure 1A). Hengduan Mountain is an important geographical boundary between the first and second steps of China, with obvious vertical zonality and dramatic changes in geomorphology and climate. Marine glaciers are developed in this area, and many glacial lakes are developed between 3,900 and 4,000 m above sea level (Zhang et al., 2012). Lake Cuoqia is lower than the forest line, with an area of 0.07 km², an average depth of 13.2 m and a maximum depth of 26 m (Figure 1B, Chai et al., 2018). The lake is hydrologically closed with no visible surface inflow and outflow. It is mainly supplied by atmospheric precipitation and ground melting snow water (Zhang et al., 2022). The vegetation around the lake is almost undisturbed by human activities, mainly subalpine low temperature coniferous trees, such as Abies and Rhododendron shrubs (Xiao et al., 2014).

The region is mainly affected by the Indian monsoon, with the same period of rain and heat. The region belongs to the temperate continental monsoon climate, with abundant solar radiation throughout the year and small annual temperature difference. According to the modern meteorological data of Shangri La Meteorological Station (27°30′0″ N, 99°25′12″ E; elevation: 3276.7 m), the nearest meteorological station to Lake Cuoqia, mean annual air temperature is 6.01°C, and mean annual precipitation is 624.72 mm. The temperature in this area is the highest in July (average monthly temperature 13.9°C) and the lowest in January (average monthly temperature - 2.3°C). The precipitation is mainly concentrated in June to September (Figure 1C). The monthly average humidity changes are between 58% (December) and 79% (August). From 1958 to 2015 AD, the mean annual precipitation has no obvious change, while the mean annual air temperature shows an obvious upward trend with 0.03°C/year (Figure 1D).

In May 2014, a pair of sediment core (CQ1 and CQ2) were obtained using Hon Kajak large-diameter (9 cm) gravity sampler in the center of the Lake Cuoqia (Figure 1B). The two cores are 37 cm and 30 cm, respectively, and composed of humus black mud. The cores were sampled at an interval of 0.5 cm in the field. The samples were stored in self-sealing bags and refrigerated at 4°C for analysis. We also collected plant samples from trees, shrubs, grasses and surface soil around the lake.

CQ1 is used for proxy analysis of brGDGTs, fatty acids and its hydrogen isotopes, n-alkanes, total organic carbon (TOC), total nitrogen (TN) (Chai et al., 2018) and elements. CQ2 is used for ²¹⁰Pb/¹³⁷Cs dating to further calibrate the age model based on the CRS (Constant Rate of ²¹⁰Pb Supply) model (Appleby and Oldfeld, 1978). The depth-age sequence of both cores was previously published in Chai et al. (2018) and Zhang et al. (2022).

About 1–3 g freeze-dried samples were extracted 4 times through ultrasonic shaker using organic solvents (dichloromethane: methanol = 9:1, v/v), ensuring complete extraction of organic matter from samples. After drying with N₂ gas, extracted total lipids were hydrolyzed using 6% KOH in Methyl alcohol solution for 12 h. Then, the supernatant was obtained after adding NaCl and n-hexane and centrifuging. Add 1.5 mL HCl (6 Mol) and 1.5 mL n-hexane to the bottle containing the sample solution to obtain the component of fatty acids. Finally, the neutral supernatant containing n-alkanes and GDGTs were further extracted through silica gel column chromatography using n-hexane and MeOH, respectively.

A total of 30 samples are determined for the fatty acids and δD values using Delta-V isotope ratio mass spectrometry (IRMS) instrument (Thermo Finnigan) via a high-temperature pyrolysis reactor at 1430°C. The instrument parameter settings and data analysis methods were referred to Liu and Liu, 2019. A total of 52 samples are analyzed for the n-alkanes via an Agilent 7,890 Gas Chromatography and the conditions for the Gas Chromatography following the previous research (Zhang et al., 2019).

A total of 68 samples are analyzed for the brGDGTs via UPLC-APCI-MS (the ACQUITY I-Class plus/Xevo TQ-S system) equipped with two coupled UPLC silica columns (BEH HILIC columns, 3.0 × 150 mm, 1.7 μm; Waters) in series, fitted with a pre-column and maintained at 30°C. The instrument can fully separate of 5- and 6-methyl isomers with improved chromatographic procedure. The samples were dissolved in 1000 μL n-hexane and injected for 4 μL for analysis. BrGDGTs were eluted at a constant flow rate of 0.4 mL/min for 80 min. The mobile phases of A and B, where A = hexane and B = hexane: isopropanol (9:1, v/v), were run isocratic ally with 82% A and 18% B for 25 min, followed by a linear gradient to 65% A and 35% B for 25–50 min, then to 100% B for 50–60 min with another 20 min re-equilibration. BrGDGTs were ionized in the APCI source at a probe temperature of 550°C, voltage corona of 5.0 μV, voltage cone of 110 V, gas flow desolvation of 1,000 L/h, gas flow cone of 150 L/h and collision gas flow of 0.15 mL/min. BrGDGTs isomers were detected using the selective ion monitoring (SIM) mode via [M + H]⁺ ions at m/z 744 for the C₄₆ standard, m/z 1,050, 1,048, 1,046, 1,036, 1,034, 1,032, 1,022, 1,020, and 1,018 for brGDGTs compounds (Hopmans et al., 2016). The modern samples were analyzed with the C₄₆ standard and the relative concentrations of brGDGTs were calculated according to the integrated peak areas. Lipid preparation, n-alkanes and brGDGTs analysis were performed at the State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Science. The fatty acids and its hydrogen isotopes were analyzed at State Key Laboratory of Loess and Quaternary Geology, Institute of Earth Environment, Chinese Academy of Sciences.

Parameters of average chain length (ACL) associated with n-alkanes and fatty acids were calculated as follow, respectively (Ficken et al., 2000; Liu and Liu, 2019).

Where the Ci is the abundance of the ith n-alkanes and fatty acids.

The site-specific calibration of Lake Cuoqia was established using a stepwise regression method between brGDGTs fractional abundance of short-core CQ1 and the instrumental temperature record during the warm season (from March to October, T_M-O; Zhang et al., 2022). Such calibration has also been verified by reconstructed temperature record of another pair core (CQ2) since 1950 AD. Thus, the equation was further used to quantitatively reconstruct the temperature sequences of both cores over the past 300 years.

Methanol correction formula is as follow (Yang and Huang, 2003):

δDn−Fas, δDn−FAMEs and δDmethanol represent value of fatty acids, fatty acid methyl ester and methanol δD, respectively. The value of δDmethanol= − 123‰.

For better discuss the driving mechanisms of climatic change, we calculated the temperature difference obtained by subtracting the average temperature of the Indian-Pacific Ocean from our reconstructed temperature. The specific method is firstly to normalize the difference between our reconstructed temperature and the temperature of the Indian-Pacific Ocean before interpolating to the same resolution.

Previous study shows that the brGDGTs in Lake Cuoqia mainly come from autochthonous sources, which are supported by multiple lines of evidence including comparison of brGDGTs distribution between surface sediments and down-core samples, ternary plots analysis (tetra-, penta-, and hexamethylated brGDGTs), relationship between the concentration of brGDGTs in surface sediments and water depth and ∑IIIa/∑IIa calculation (Zhang et al., 2022). The brGDGTs can well capture the temperature changes during the instrumental period at Lake Cuoqia with high correlation (R² = 0.89) to nearby meteorological data (Zhang et al., 2022). Using the same correction equation, we further quantitatively reconstructed the temperature changes in the past 300 years. The reconstructed temperature dropped continuously before 1950 AD and rose rapidly afterwards (Figure 2A). BIT (Branched Isoprenoid Tetraether index) values are the ratio of branched GDGTs to all fractional abundance of GDGTs (including branched and isoprenoid tetraether) and have been widely used as a proxy to evaluate the stability of the sedimentary environment (Zhang et al., 2019, 2022; Yan et al., 2021; Zhao et al., 2021a). The change of the BIT values was quite stable varying from 0.95 to 1 (Figure 2B), indicating stable sedimentary environment and the applicability of the calibration to the whole core. Our temperature results are supported by warm-season temperature (from March to October) from regional meteorological station data during 1958–2015 AD for both long-term trend and amplitude of variation (Figure 2A).

The reconstructed temperature shows consistent with other proxies from the same core such as ACL values of both n-alkanes and fatty acids, TOC and TN (Figures 2C, D). Previous study shows that the mid- and long-chains of leaf wax mainly from terrestrial plants, which is sensitive to temperature changes and can be used as an indicator of temperature changes (Zhou et al., 2005). Although the overall pattern of temperature changes is consistent, the slight discrepancy between them is possibly due to different responses to climate change. TOC and TN are two fundamental proxies for describing organic matter content in sediments, mainly reflecting the primary production of biomass which is related with regional climate changes (Meyers and Ishiwatari, 1993). The decline of TOC and TN may indicate the gradually cold-dry climate conditions, in accord with the variation of our reconstructed temperature.

Our reconstructed temperature is also consistent with previously limited regional temperature records (Duan and Zhang, 2014; Zhang et al., 2017; Wu et al., 2018). For instance, July temperature based on subfossil chironomids from Lake Tiancai showed an overall decrease with a rapid increase after 1970 AD, albeit with an abnormal value at 1950 AD and quite low resolution (Figure 3B; Zhang et al., 2017). Our absolute temperature has lower values than Lake Tiancai for the past 300 years, which can attribute to differences of reconstructed season and elevation (higher ~60 m). Moreover, the similar long-term trend can also be observed from pollen-based July temperature record at the Lake Xingyun (Figure 3C; Wu et al., 2018). The higher-resolution warm-season (from April to September) temperature from tree ring showed a slight decrease trend before 1920 AD with a reverse afterwards (Figure 3D; Duan and Zhang, 2014). The low temperatures centered at 1870 AD and 1980 AD corresponded with our reconstructed temperature (Figure 3D). The mismatch in the warming time may be attributed to dating uncertainty. It is worth noting that our temperature shows obvious discrepancy with the trend of mean annual air temperature reconstructed by brGDGTs from Lake Tiancai and sea surface temperature from Indian-Pacific Ocean (Tierney et al., 2015; Feng et al., 2019), both showing a continuous warming (Figures 3E, F). This may be attributed to seasonal difference between warm-season and mean annual temperature, which is confirmed to be present at longer Holocene scales (Sun et al., 2021; Zhang et al., 2022). In conclusion, our reconstructed 300 years quantitative temperature is reliable and agrees well with regional limited temperature records.

Previous studies suggest that the hydrogen isotopes of fatty acids and n-alkanes from terrestrial plants can well record the isotopic changes of atmospheric precipitation (Eglinton and Eglinton, 2008; Sachse et al., 2012). δD of C₂₂ from fatty acids has similar fraction process of n-alkanes δD from precipitation isotope in hydrological cycles (Hou et al., 2006; Contreras-Rosales et al., 2014). However, the controlled factors of leaf wax δD include local rainfall, soil evaporation, vegetation fractionations, etc. (Dansgaard, 1964; Cai et al., 2012; Sachse et al., 2012; Zhang et al., 2020). The water required for plants in the lake catchment mainly derived from soil water which is influenced by monsoon precipitation and soil evaporation effect (Sachse et al., 2004, 2012; Zhao et al., 2021b). Thus, leaf wax δD should mainly reflect the variations in the relative humidity. The isotopic fractionations may also exist during lipid biosynthesis in plant, and possible evapotranspiration between soil and lipid leaf wax water (Sachse et al., 2012). Some studies from southwestern China demonstrate that the isotope (δD and δ¹⁸O) of tree ring indeed indicates the changes of relative humidity on centennial time scale (An et al., 2014), and has been verified by the regional instrumental data. Note that the effect of vegetation fractionation also exists in tree-ring δD with little influence, similar to our leaf wax δD of fatty acid (An et al., 2014). Therefore, our isotope records the changes of regional relative humidity with positive δD of C₂₂ indicating dry environment, and vice versa.

In the past 300 years, the gradually enriched δD of C₂₂ from the Lake Cuoqia indicates a continuously dry condition, which shows good relation with the relative humidity measured by instrument data over the past decades (Figure 2E). Our δD-based relative humidity is also consistent with the ratio of Fe/Mn from the same core (Figure 4B). Fe/Mn can indicate redox state and further indicate the rise and fall of lake level with high ratio of Fe/Mn corresponding to high lake level, and vice versa (Mackereth, 1966). Similar changes can also be observed in another proxy of Rb/Sr. ratio in the same core (Figure 2G), which is widely used to reflect the intensity of chemical weathering with low values for intense chemical weathering related to humid environment, and vice versa (Chen et al., 2008; Liu J. et al., 2014).

The consistent changes of the relative humidity can also be recorded by other geological archives including tree ring, stalagmite and lake (An et al., 2014; Xiao et al., 2014; Tan et al., 2018; Xu et al., 2018, 2019). For example, the relative humidity, reconstructed by high-resolution tree ring δ¹⁸O from Batang-Litang Plateau (BLP) of southeastern Tibetan Plateau (Figure 1A), showed a long-term drying trend in the past 300 years (Figure 4C; An et al., 2014). Also, similar trend can be observed in another δ¹⁸O of tree ring from low-latitude highlands (LLH) of southwestern China (Figure 1A), revealing an apparent drying trend especially after 1840 AD (Figure 4D; Xu et al., 2018, 2019), which can be supported by the reconstructed cloud cover records using composite δ¹⁸O of three tree-ring chronologies from southeastern Tibetan Plateau (Figure 4E; Liu X. et al., 2014). Similarly, many δ¹⁸O records of stalagmite in nearby regions also show consistent changes with our reconstructed relatively humidity. For instance, the precipitation index and δ¹⁸O of stalagmite from Shenqi Cave (SQ) in southwestern China (Figures 4F, G; Tan et al., 2018) showed a persistent positive trend, indicating a drying environment. In addition, the gradually drying environment is also supported by pollen data in sediments of Lake Tiancai, in which the tree pollen of Tsuga gradually decreases (Xiao et al., 2014). In summary, the reconstructed relative humidity is consistent with proxies from same core and is supported by instrumental data and regional precipitation/relative humidity records.

Our reconstructed temperature and relative humidity showed consistent changes between 1700 AD and 1950 AD toward to gradually cold-dry trend, whereas started to decouple after 1950 AD, manifested as increasing temperature and decreasing relative humidity (Figures 2A,E). The combined pattern of reconstructed temperature and relative humidity is characterized with decoupling at 1950 AD, when the reason has not yet completely recorded and discussed by regional archives. Here, we focus on analyzing and explaining the underlying causes and driving factors of decoupling before and after 1950 AD.

The continuous decrease of temperature and relative humidity before 1950 AD was in accord with the decreasing warm-season insolation (from March to October) at 26° N (Figure 3H; Laskar et al., 2004; Sun et al., 2021). The decreasing insolation reduced total energy received at the earth surface, resulting into the decline in regional temperature. The decreasing insolation also weakened the intensity of Indian summer monsoon and further reduced the precipitation and/or relative humidity in our study area (Tan et al., 2018). Moreover, the persistent decline of relative humidity may be related to the decreasing thermal contrast between sea surface temperatures of the tropical Indian-Pacific Ocean and land temperature in our region (Figure 3G), which determines the intensity of water-vapor transports dominated by the Indian summer monsoon (Bansod et al., 2003; Feng and Hu, 2005; An et al., 2014). Furthermore, the pressure difference between Tibetan Plateau and tropical Ocean may also affect the monsoon precipitation and relative humidity in the southeastern Tibetan Plateau (Rashid et al., 2011). Previous studies suggest that the years with high relative humidity are related to the low-pressure conditions on the southeastern Tibetan Plateau, while the pressure field on the Indian Ocean is opposite (An et al., 2014). When the Tibetan Plateau maintains a high-pressure ridge in summer, the intensity of Indian summer monsoon weakens, reducing the movement of ocean air mass from the Indian Ocean to the plateau (Charles et al., 1997; Xu et al., 2009). Therefore, the path of rain storms will move southward, resulting in low relative humidity. In addition, the decreasing relative humidity over the past 300 years has high coherence with overall southward shift of Intertropical Convergence Zone (Figure 4H; Tan et al., 2019) and intensified EI Niño-like conditions (Figure 4I; Man and Zhou, 2011), indicating a pivotal role of low-latitude driving force to southeastern Tibetan Plateau.

After 1950, our reconstructed temperature record showed consistent changes to the rapid increase in greenhouse gases emission caused by human activity (Figure 3I), indicating a close connection between them (Crowley, 2000). Although increased temperature can lead to more water-vapor supply and larger temperature difference between sea and land, the relative humidity showed an overall decrease trend during this period (Figure 4A). The decreased relative humidity may be caused by enhanced evaporation associated with unprecedented warming. In addition, the decreasing relative humidity was possibly related to the aerosol-affected Anthropocene warming as well which can lead to a weakening of summer monsoon intensity and thus result into dry environment (Liu et al., 2017). In the future, the continued and rapid warming would further decrease the relative humidity, and more attention should be taken for extreme climate changes in the Tibetan Plateau region.