1 Introduction
Global warming has generated major concerns about the trend and amplitude of future temperature changes. While climate predictions rely mainly on climate models, quantitative paleotemperature reconstructions can help improve the accuracy of climate simulations by providing tests and evaluations of climate models (
Snyder, 2010;
Braconnot et al., 2012). The late Holocene is a critical period in the Earth’s climate history because of the transition from forcing solely by natural factors to the additional forcing of anthropogenic greenhouse gases (
IPCC, 2013). Quantitative reconstructions of temperature changes for the late Holocene are crucial for assessing the significance of recent anthropogenic warming within the context of natural climate variability (
Feng et al., 2019;
Yao et al., 2019;
Li et al., 2023a).
Currently, the rate of temperature increase at high-elevation sites is much higher than the global average (
Bradley et al., 2006; Mountain Research Initiative EDW Working Group, 2015). The Tibetan Plateau (TP), covering 2.5 million km
2 and with an average elevation of > 4000 m, is the largest high-elevation region on the Earth and is especially sensitive to global environmental changes (
Yao et al., 2012,
2019;
Li et al., 2021a). Observation data from meteorological stations on the TP indicate surface warming at a rate twice that observed globally (
Chen et al., 2015). The TP is often referred to as ‘Asia’s Water Tower’, as it hosts numerous mountain glaciers that provide an indirect water supply for hundreds of millions of people throughout Asia (
Immerzeel et al., 2010). Over the last few decades, glaciers on the TP have experienced rapid and pronounced changes (
Yao et al., 2019), and there are increasing concerns about the potential effects on the hydrology, ecosystems, and human population of this region and elsewhere. The retreat and advance of the TP glaciers are closely related to temperature variations (
Bolch et al., 2012;
Jacob et al., 2012); hence, knowledge of paleotemperature variations on the TP can provide a valuable context for assessing the current and possible future status of its glaciers.
There are numerous paleotemperature reconstructions since the last deglaciation for the TP and the adjacent areas (
Wang et al., 2015a;
Hou et al., 2016;
Li et al., 2017;
Zhang et al., 2017;
He et al., 2020;
Zhao et al., 2021;
Sun et al., 2022;
Zhang et al., 2022). These records generally indicate a cooling trend of summer temperatures during the Holocene, although the patterns of mean annual temperature (MAAT) change on the TP remain controversial (
Chen et al., 2020;
Sun et al., 2022). In particular, MAAT reconstructions across the TP during the middle to late Holocene have yielded conflicting results. Several of these MAAT reconstructions show a cooling trend since the middle Holocene (
Zhao et al., 2021), while others reveal a warming trend (
Thompson et al., 1997;
Li et al., 2017;
Sun et al., 2022;
Zhang et al., 2022). The mechanism responsible for these different temperature trends remains unclear. Hence, more quantitative MAAT reconstruction across the TP are needed to better understand the Holocene temperature dynamics of this region.
Branched GDGTs (brGDGTs) are increasingly regarded as a promising tool for continental climate reconstruction (
Weijers et al., 2007;
Li et al., 2023b). brGDGTs are membrane lipids synthesized by heterotrophic bacteria that contain two C
28 alkyl chains with 4–6 methyl substituents and 0–2 cyclopenthy moieties, and they are ubiquitous within a wide range of terrestrial depositional environments (
Weijers et al., 2006;
Schouten et al., 2013). The brGDGTs-producing bacteria can alter the fluidity of the lipid membrane, enabling it to adapt to changes in the environment, which are reflected in the quantity of methyl and cyclopentane moieties (
Weijers et al., 2007). Based on a survey of global soils, the degree of methylation (MBT) has been shown to depend primarily on MAAT and soil pH, while the cyclization ratio (CBT) is predominantly related to soil pH (
Weijers et al., 2007). Therefore, MBT/CBT and its modified version (e.g., MBT′/CBT) have been established to reconstruct paleotemperatures (
Weijers et al., 2007;
Peterse et al., 2012). More recently, improved chromatographic techniques have enabled the separation of 5- and 6-methyl brGDGTs, resulting in a set of new brGDGTs proxies, which were used to recalibrate traditionally defined MBT–CBT indices (
De Jonge et al., 2014;
Yang et al., 2015). Investigations of brGDGTs distributions in lake sediments from sites on the TP have shown that brGDGTs proxies are a reliable means of reconstructing past temperature variability across the TP (
Wang et al., 2016;
Liang et al., 2022). Specifically, brGDGTs have been successfully applied to TP lake sediments to reconstruct palaeotemperature variations during the Holocene (
Li et al., 2017;
Feng et al., 2019;
Li and Fan, 2019;
Wang et al., 2021;
Sun et al., 2022;
Zhang et al., 2022).
In this study, we obtained a brGDGTs-based paleotemperature record spanning the past 4700 years from Xiada Co on the western TP. We have already reconstructed a 2000-year-long temperature record for Xiada Co using the brGDGTs-based temperature proxy (Index1) (
Li et al., 2019). However, as a regional calibration of brGDGTs for TP lakes using the improved separation technique was not available at that time, the ~2000-year-long temperature record from Xiada Co presented in
Li et al. (2019) was calculated used a global calibration. However, considering the unique climate and environment of the TP, a regional rather than a global calibration for brGDGTs-based palaeothermometry is clearly needed (
Wang et al., 2021). Recently,
Liang et al. (2022) analyzed 29 surface sediment samples from lakes in TP using the improved separation technique. They found that all temperature variables had a stronger linear relationship with the degrees of methylation of 6-methyl brGDGTs (MBT′
6Me) than 5-methyl brGDGTs (MBT′
5Me), which was also found in a survey of the distribution of brGDGTs in 35 lakes across China (
Dang et al., 2018). Subsequently, they produced a new MBT′
6Me-MAAT calibration based on the 29 surface sediment samples from lakes on the TP, together with 39 additional lake surface sediments from China (
Dang et al., 2018). In the present study, we applied the new calibration to a sediment core spanning the last ~5000 years from Xiada Co. Our objectives were to document MAAT variations in the western TP over the past ~5000 years, and to determine the factors that contribute to the spatiotemporal differences of paleotemperature variations across the TP.
2 Study site
Xiada Co (33°23′N, 79°21′E, 4358 m a.s.l) is a small freshwater lake located on the western TP, ~50 km west of the town of Rutog (Fig.1). The lake is fed mainly by precipitation and glacier meltwater. The results of a modern lake survey indicated that the salinity of the surface lake water is 0.15 g/L, and the pH of the lake water is 8.6. The lake has a surface area of 8 km
2 with a maximum depth of 20 m (
Li et al., 2019). A long-term instrumental data series from Shiquanhe meteorological station, ~120 km south-east of Xiada Co, shows that the MAAT was 1.02°C over the 30-year interval of 1981–2010 CE (
Wang et al., 2021). The current precipitation in this region is monsoonal, and the summer rainfall accounts for > 80% of the total annual precipitation (
Li et al., 2019).
3 Materials and methods
3.1 Sampling
A 273-cm-long sediment core (XDC2014-1), was collected from the center of Xiada Co at a water depth of 19 m using a Uwitec piston corer (Fig.1). The upper part of the sediment core was continuously subsampled at an interval of 0.5 cm, and below 100 cm the core was subsampled at an interval of 1 cm. All samples were stored in a freezer prior to laboratory analysis (
Li et al., 2019).
3.2 Lipid extraction and brGDGTs analysis
brGDGTs were extracted from the lake sediments using the methods described by Li et al. (2019). Freeze-dried samples (~5 g) were ultrasonically extracted with dichloromethane/methanol (9:1, v/v) (15 min, 3 ×, 30°C). The total extract was chromatographed using an activated Al2O3 column (for 2 h at 150°C). The nonpolar and polar fractions (the latter containing GDGTs) were eluted using n-hexane/dichloromethane (9:1, v/v) and dichloromethane/methanol (1:1, v/v), respectively. The extracts were concentrated in as stream of nitrogen gas and then dissolved in n-hexane/isopropanol (99:1, v/v). A 0.2 μm PTFE filter was used to remove large molecular compounds and particulate matter prior to HPLC-MS analysis.
GDGTs analysis was performed using HPLC-APCI-MS (Agilent 1260 HPLC system with 6100 MS), with autoinjection, at the Institute of Tibetan Plateau Research, Chinese Academy of Sciences (ITPCAS). Separation of 5- and 6-methyl brGDGTs isomers was achieved with three Hypersil Gold Silica LC columns in sequence (each with dimensions of 100 mm × 2.1 mm, 1.9 μm; Thermo Fisher Scientific, USA). The instrumental setups for HPLC were as follows: column temperature 40°C, injection volume 10 μL, flow rate 0.2 mL/min. GDGTs were eluted isocratically with 84% A and 16% B for the first 5 min, where A is
n-hexane and B is EtOA, followed by a linear gradient change to 82% A and 18% B from 5 to 65 min and then to 100% B for 21 min, followed by 100% B for 4 min to wash the column, and then back to 84% A and 16% B to equilibrate the column. APCI-MS conditions were as follows: nebulizer pressure 60 psi, vaporizer temperature 400°C, drying gas flow rate 6 L/min and temperature 200°C, capillary voltage 3500 V, corona 5 μA. Detection was performed in selected ion monitoring (SIM) mode via [M + H]
+ at m/z 744 for the C
46 standard, 1302, 1300, 1298, 1296, and 1292 for iGDGTs and 1050, 1048, 1046, 1036, 1034, 1032, 1022, 1020, 1018 for brGDGTs. GDGTs were quantified using an external standard, and the peak areas were manually integrated, assuming an identical response factor for the GDGTs. The definition of the structure and nomenclature of brGDGTs follows
Yang et al. (2015).
4 Results and discussion
4.1 Chronology
The chronology of core XDC2014-1 was constructed using
210Pb and
137Cs analyses of the upper 10 cm (Fig.2(a)), together with 6 AMS
14C ages of total organic carbon from the lower part of the core (Fig.2(b)). The age-depth model of core XDC2014-1 is discussed in detail in Li et al. (
2019). Briefly,
210Pb and
137Cs analyses were conducted by gamma spectrometry, using a high-purity well-type germanium detector (ORTEC GWL-120–15), at the Institute of Tibetan Plateau Research, Chinese Academy of Sciences. The age-depth relationship of the upper 10 cm of the core, together with the dating uncertainty, was established using the constant rate of supply (CRS) model (
Appleby and Oldfield, 1978). A
137Cs peak occurs at the depth of 4.5 cm, likely corresponding to the Chernobyl nuclear accident in 1986 CE, which is consistent with the
210Pb dating results. The final chronology for the most recent sediments is based on
210Pb dating (Fig.2(a)), according to which the upper 10 cm of the core represent the interval of 1919–2013 CE, corresponding to a sedimentation rate of ~0.1 cm/yr.
Because of the lack of terrestrial macrofossils in core XDC2014-1 the AMS ages were obtained from bulk organic carbon. The
14C reservoir ages of lakes on the TP show significant temporal variability, and they can vary by as much as ~5000 years within a single lake between the last deglaciation and the late Holocene, resulting in large age uncertainties in reservoir age corrections (
Hou et al., 2012). In addition, large errors may be introduced by assuming a constant sediment accumulation rate if the profile spans significant environmental changes, which can have a major impact on calculating a reservoir effect (
Zhou et al., 2014). To reduce the chronological uncertainty, it is advisable to consider the possible temporal changes of
14C reservoir ages and to determine the reservoir age for different sections of the sediment core (
Zhou et al., 2014). The differences in the total organic content (TOC) of core XDC2014-2 indicate an abrupt change in the sedimentation rate at the depth of 80 cm. Therefore, to calculate the reservoir ages we divided the core into two sections, with the boundary at 80 cm. All the
14C data were first calibrated to calendar ages using CALIB 7.0.2. Linear regression of the uppermost two calibrated
14C ages yielded an age of 2972 cal yr BP for the depth of 0 cm, which can be regarded as the average reservoir age for the upper section of the core. According to the
210Pb dating results, the depth of 8 cm has the age of 0 yr BP (1950 CE), while 3038 cal yr BP is inferred from the extrapolation of the upper two radiocarbon dates. Thus, 3038 years can also be regarded as the reservoir age at that depth. We used the mean value for the two reservoir ages (3005 years) as the reservoir age for the upper section of the core. Subsequently, a linear fit was applied to the lower four calibrated
14C ages which yielded a
14C age of 4199 cal yr BP for the depth of 80 cm, while the reservoir age–corrected
14C age at the depth of 80 cm was 634 cal yr BP. Therefore, we took the difference (3565 years) between the two ages as the reservoir age below 80 cm. All the reservoir ages were then subtracted from the original calibrated ages and the final chronology for the sediment core was constructed based on the 6 reservoir age–corrected
14C ages together with the
210Pb/
137Cs chronology for the upper 10 cm. Linear interpolation between the reservoir age-corrected ages was used to obtain the ages of all the sediment subsamples (
Wu et al., 2013).
4.2 Distribution and source of brGDGTs in the sediments of Xiada Co
The 5- and 6-methyl brGDGTs in all samples were chromatographically separated. Non-cyclopentane moieties (IIIa, IIIa′, IIa, IIa′, and Ia) dominate the distribution of brGDGTs, contributing 83% of the total amount within the core samples (Fig.3(a)). The relative abundances of brGDGTs containing 1–2 cyclopentane moieties were relatively low and sometimes too low to detect. The most abundant brGDGTs are pentamethylated brGDGTs (41%), followed by hexamethylated (38%) and tetramethylated (21%) brGDGTs.
To further evaluate the applicability of brGDGTs-based proxies in Xiada Co, it is necessary to determine the source of the brGDGTs within the lake (
Naeher et al., 2014). Hence, we compared the brGDGTs in the sediments from Xiada Co with published data from Tibetan soils (
Ding et al., 2015) and lake sediments (
Liang et al., 2022). The results demonstrate that the non-cyclopentane moieties (IIIa, IIIa′, IIa, IIa′, and Ia) dominate the distribution of brGDGTs in all samples (Fig.3(a)). In addition, all are dominated by pentamethylated brGDGTs, followed by hexamethylated brGDGTs and tetramethylated brGDGTs. Nevertheless, there are subtle differences; for example, the brGDGTs in the sediments samples from Xiada Co and the sediments of other Tibetan Plateau lakes have a lower fractional abundance of pentamethlyated brGDGTs (41% for Xiada Co and 44% for the other Tibetan lake sediments) relative to the Tibetan Plateau soils (54%), but a higher fractional abundance of hexamethylated brGDGTs (38% for Xiada Co down-core and 39% for the other Tibetan lake sediments) relative to the Tibetan soils (27%). This indicates that the brGDGTs in the core sediments from Xiada Co are generally of
in situ origin and that the terrestrial input if brGDGTs is negligible. Also, the ternary diagram of tetra-, penta-, and hexa-methylated brGDGTs can separate soil-derived brGDGTs from those of in situ (aquatic) origin (
Sinninghe Damsté, 2016). In a ternary plot, the data from Xiada Co are distributed close to those of other Tibetan lake sediments, but they are distinguished from Tibetan soils (
Ding et al., 2015) and global soils (
Naafs et al., 2017) (Fig.3(b)). This further indicates that that the brGDGTs from Xiada Co are mainly produced within the water column.
4.3 Late Holocene temperature reconstruction for Xiada Co
We reconstructed the temperature variations in Xiada Co using both soil-based (
Weijers et al., 2007;
Peterse et al., 2012;
De Jonge et al., 2014;
Ding et al., 2015) and lake-specific (
Günther et al., 2014;
Wang et al., 2016;
Liang et al., 2022) functions on global and regional scales. The formulas used to calculate the brGDGTs indices and calibrations are shown in Tab.1. Generally, the temperature reconstructions show a similar trend of variation but with different absolute values (Fig.4). Among these calibrations, that of
Liang et al. (2022) appears to be the most reliable because the current temperature calculated by this function (1.5°C) is close to the MAAT from 2011 to 2014 at Shiquanhe Meteorological station (1.6°C). Additionally, comparison of the MBT′
6Me-reconstructed MAAT with the nearby meteorological data set spanning the past 37 years indicates a significant correlation between the two, both showing a minimum at ∼2000 CE and relatively high MAAT before and after 2000 CE (
Liang et al., 2022). In addition, this calibration provides the most extensive coverage of Tibetan lakes (31 lakes), increasing our confidence in applying it to lakes on the TP. Therefore, we adopted the MBT′
6Me -MAAT relationship developed by
Liang et al. (2022) to convert the data from Xiada Co to MAAT.
As shown in Fig.4(a), the brGDGTs-based MAAT record from Xiada Co has an average of ~2.8°C, with the range of around –2.1°C–13.0°C over the past 4700 years. Specifically, the MAAT shows a general cooling trend during the past 4700 years. A relatively warm interval is evident during ~4700–2200 cal yr BP, with MAAT fluctuating between 5.3°C and 13.0°C. The MAAT then decreased by ~4.4°C at ~2100 cal yr BP, and it continued to decrease to the present-day with centennial-scale oscillations superimposed, centered at ~800 cal yr BP, ~600 cal yr BP, and ~190–170 cal yr BP. MAAT decreased abruptly at ~500–300 cal yr BP, reaching the lowest average annual temperature in the lake over the past 4700 years.
Several researchers have proposed a temporal consistency between climatic events and the evolution of human civilization, implying a causal relationship (
deMenocal, 2001;
Zhang et al., 2010;
Kennett et al., 2012). Furthermore, several studies have suggested that climate change played a role in the transitions between dynasties and the associated societal evolution in China (
Yancheva et al., 2007). Thus, high-resolution climate records are potentially valuable for improving our understanding of the impacts of past climate change on human societies. Xiada Co is located near the ruins of the Guge Kingdom, which was established in the mid-10th century in the western TP and flourished for ~700 years (
Kathayat et al., 2017). However, the Guge Kingdom abruptly collapsed in the 17th century, which was ascribed to an aridification trend caused by a monsoonal minimum (
Cheng et al., 2010;
Kathayat et al., 2017;
Li et al., 2019). The brGDGTs-based temperature record from Xiada Co suggests that the temperature decreased abruptly at ~500–300 cal yr. The decreased crop yield potentially caused by this temperature decrease may have contributed to the collapse of the Guge Kingdom (
Liang et al., 2022). Additionally, this temperature decrease was synchronous with a major decline in the human population reflected by changes in the absolute concentrations of fecal stanols in core XDC2014-1 from Xiada Co (
Li et al., 2023c), which also indicates that abrupt climate change may have contributed to the collapse of the Guge Kingdom. However, possible errors and uncertainties associated with the
14C chronology for core XDC2014-1 necessitate caution in establishing a correlation between temperature changes and the evolution of regional civilizations, and further research is needed to constrain the sedimentary chronology for Xiada Co.
4.4 Pattern of temperature changes on the TP over the past 5000 years
Our reconstructed MAAT record from Xiada Co indicates a cooling trend with an abrupt temperature decrease at ~500–300 cal yr BP (Fig.5(a)). To further evaluate our record, we compared it with other paleoclimate records from the TP and adjacent areas. The criteria we used for selecting these records was that they were quantitative and were interpreted by the authors as having a clear seasonal significance.
Changes in the MAAT on the TP during the past 5000 years have been reconstructed using various temperature proxies, including pollen assemblages (
e.g., Chen et al., 2020), ice core δ
18O (
e.g., Thompson et al., 1997;
Pang et al., 2020), and brGDGTs (
e.g., Li et al., 2017;
He et al., 2020;
Wang et al., 2021;
Zhao et al., 2021). Unfortunately, some of these reconstructions are mutually inconsistent (Fig.5). A cooling trend over the past ~5000 years has been reported by several recent studies; for example, Chen et al. (
2020) compiled a set of pollen data for the TP (Fig.5(e)) and found that the annual temperature gradually decreased since the middle Holocene. In addition, the temperature anomalies reconstructed from the δ
18O record of the Chongce ice core, which has a slight winter bias, show a cooling trend during the past 5000 years (Fig.5(f)) (
Pang et al., 2020). Likewise, brGDGTs records from Tengchongqinghai lake (Fig.5(b)) and Lugu lake (Fig.5(c)) (
Zhao et al., 2021), in the Hengduan Mountains on the south-eastern margin of the TP, also depict cooling trend from the middle Holocene. The similarity of the temperature records from Xiada Co (Fig.5(a)), Tengchongqinghai lake (Fig.5(b)) and Lugu lake (Fig.5(c)) suggest that MAAT variations were synchronous in the western and south-eastern TP, and compatible with the global trend indicated by proxy-based paleotemperature syntheses (Fig.5(d)) (
Marcott et al., 2013). However, the ice core δ
18O record from the Gulliya ice core shows a long-term warming trend during the middle to late Holocene (Fig.5(k)) (
Thompson et al., 1997), which differs from our MAAT record and from other TP temperature records, including those from the Chongce ice cap (Fig.5(f)), Tengchongqinghai lake (Fig.5(b)) and Lugu lake (Fig.5(c)). Furthermore, the brGDGTs-based MAAT records from Bangong Co (Fig.5(l)) (
Wang et al., 2021) and Aweng Co (Fig.5(i)) (
Li et al., 2017) on the western margin of the TP; from Ngamring Co on the southern TP (Fig.5(h)) (
Sun et al., 2022); and from the Hongyuan peatland on the eastern TP (Fig.5(m)) (
Yan et al., 2021) all demonstrate a warming trend during the past 5000 years. In addition, beyond the TP, a peat record from the Altai Mountains (Fig.5(j)) also shows distinct temperature increase during the past 5000 years (
Wu et al., 2020). The warming trend since ~5000 years cal yr BP reflected by these temperature records is completely at variance with our brGDGTs-based temperature reconstruction from Xiada Co. It should be noted that brGDGTs-based MAAT records do not consistently show a prominent warming or cooling trend during the past ~5000 years. For example, although the MAAT record from Lingge Co on the central TP fluctuated sharply during the past 5000 years, there was only a weak overall warming trend (
He et al., 2020). Furthermore, the ice-free-season temperature (from March to October, T
M-O) based on brGDGTs from Cuoqia Lake (Fig.5(g)), in the hinterland of the Hengduan Mountains on the south-eastern margin of TP, shows a relatively stable pattern during 5000–1500 cal yr BP (Fig.5(g)) (
Zhang et al., 2022), which differs from the record from Xiada Co and from other brGDGTs-based temperature records.
Summer temperature changes have been reconstructed from chironomids (
Chang et al., 2017;
Zhang et al., 2017), pollen (
Chen et al., 2020), and alkenones (
Wang et al., 2015b;
Hou et al., 2016). Our brGDGTs-based temperature record from Xiada Co (Fig.6(a)) demonstrates a similar trend to the pollen–based summer temperature record (Fig.6(b)) (
Chen et al., 2020), displaying a gradual cooling trend during the past 5000 years. Likewise, the summer temperature variability inferred from subfossil chironomid assemblages from Heihai Lake (Fig.6(c)) and Tiancai Lake (Fig.6(g)), on the south-eastern margin of the TP, shows a generally decreasing trend for the past 5000 years, suggesting that summer temperatures on the south-eastern margin of the TP primarily responded to changes in the Asian summer monsoon, on the millennial scale (
Chang et al., 2017). Similarly,
Chen et al. (2020) summarized the activity of Holocene glaciers based on the published
10Be ages of moraines on the TP and in the surrounding mountains. The increased activity of Tibetan glaciers since ~5000 cal yr BP suggests that glacier accumulation occurred in response to decreasing summer temperatures (Fig.6(e)). Moreover, the δ
18O values of authigenic carbonate at Guozha Co gradually became enriched from the middle Holocene onward, reaching a maximum during 1500–500 cal yr BP (Fig.6(d)), which indicates a decrease in glacier meltwater influx to the lake since the middle Holocene, which can be attributed to decreasing summer temperatures (
Li et al., 2021b). It should be mentioned that the alkenone-based summer temperature record from Lake Qinghai shows an interval of sustained summer temperature decrease during 5000–3500 cal yr BP, after which summer temperatures gradually increased to the current level, with several fluctuations superimposed (Fig.6(f)) (
Hou et al., 2016), which differs from our record from Xiada Co and from other summer temperature records.
In summary, although the absolute values are different, the trends of the temperature record from Xiada Co is consistent with most summer temperature records (
Chang et al., 2017;
Zhang et al., 2017;
Chen et al., 2020;
Li et al., 2021b) and with some of the MAAT records (
Chen et al., 2020;
Pang et al., 2020;
Zhao et al., 2021) from the TP—all demonstrating a cooling trend during the past 5000 years. However, several other MAAT records from the TP contrast with our temperature records, showing a marked warming trend since the middle Holocene (
Thompson et al., 2006;
Li et al., 2017;
Wang et al., 2021;
Yan et al., 2021;
Sun et al., 2022). This discrepancy may partly reflect the spatiotemporal complexity of temperature variations on the TP (
He et al., 2020). Furthermore, the climate series collected from different regions of the TP are derived from multiple proxy types, each with specific principles, procedures, and potentially unique qualities. Thus, the biases inherent in each proxy archive, such as seasonality and chronological error, may be a major source of inconsistency in the reconstructed temperature series.
4.5 Possible causes of the spatial heterogeneity of temperature changes on the TP
Evaluating these paleotemperature reconstructions for the TP for the past 5000 years is potentially complicated. The discrepancies between these temperature records could be related to seasonality effects on the temperature proxies (
Liu et al., 2014;
Hou et al., 2019a;
Wang et al., 2021;
Feng et al., 2022). Many temperature proxies (such as brGDGTs, pollen, and alkenones) are controlled by seasonal processes and reflect the climate of the growing season (
Chu et al., 2012). Therefore, seasonal biases exist for these indicators and thus for the resulting temperature reconstructions. This seasonal bias may be especially pronounced in the arid, high-elevation TP region, where the modern climate is characterized by strong temperature seasonality with a long cold season and a short mild season (
Li et al., 2011). For example, the chironomid-based summer temperature records from Tiancai Lake (
Zhang et al., 2017) and Heihai Lake (
Chang et al., 2017) show a gradual cooling trend during the past 5000 years Similar characteristics are shown by a pollen-based summer temperature synthesis for the TP (
Chen et al., 2020). By contrast, the brGDGTs-based MAAT records from Bangong Co (
Wang et al., 2021), Aweng Co (
Li et al., 2017), Ngamring Co (
Sun et al., 2022), and the Hongyuan peatland on the TP (
Yan et al., 2021) all show a warming trend during the past 5000 years.
Modern investigations have confirmed the validity of brGDGTs as an MAAT proxy for both soil and lakes in the TP (
Günther et al., 2014;
Ding et al., 2015;
Wang et al., 2016;
Liang et al., 2022), and brGDGTs-based proxies have been increasingly used for MAAT reconstruction from lake sediments on the TP (
Li et al., 2017,
2019;
Feng et al., 2019;
He et al., 2020;
Wang et al., 2021;
Zhao et al., 2021;
Sun et al., 2022). However, brGDGTs proxy estimates are sometimes assumed to have a seasonal bias (
Sun et al., 2011;
Deng et al., 2016;
Foster et al., 2016;
Dang et al., 2018;
Qian et al., 2019;
Zhu et al., 2021). An investigation of alkaline lake sediments from cold regions in China has verified a warm-season bias in brGDGTs production (
Dang et al., 2018), which is consistent with numerous other studies indicating a bias toward the warm months in mid- to high-latitude lakes (
Sun et al., 2011;
Shanahan et al., 2013;
Foster et al., 2016). Recent studies suggest that whether the lakes freeze and the duration of ice cover can affect the seasonality of the reconstructed temperature (
Zhu et al., 2021;
Zhang et al., 2022). An analysis of brGDGTs from Sihailongwan maar lake (Lake SHL) in NE China showed that the brGDGTs are produced predominantly in summer and autumn due to the long duration of ice cover, from the middle of November to the end of April (
Zhu et al., 2021). Likewise, a recent study suggested that the brGDGTs in Cuoqia Lake on the TP reflect the temperature of the ice-free-season (from March to October, T
M-O), with freezing occurring during the cold season (
Zhang et al., 2022). For Xiada Co, located in a higher altitude area of the western TP, mean temperatures above freezing occur consecutively during the five months from May to September, as indicated by observations from the nearby Shiquanhe station. Thus, the brGDGTs-reconstructed annual temperature record from at Xiada Co predominantly reflects air temperature changes during the ice-free season from May to September, suggesting that a possible warm season bias exists in our reconstruction. This may partly explain why the trend of the Xiada Co temperature record is consistent with most of the summer temperature records from the TP (
Chang et al., 2017;
Zhang et al., 2017;
Chen et al., 2020;
Li et al., 2021b).
However, Drotz et al. (
2010) proposed that microbes are able to maintain both catabolic and anabolic processes under freezing conditions, with continuing brGDGTs production occurring even in cold conditions. Moreover, a study of two soil sites in China with contrasting temperature seasonality failed to detect a significant bias toward summer temperature in brGDGTs (
Lei et al., 2016). Additionally, no seasonal trends were observed in the concentration and distribution of the branched GDGTs in mid-latitude soils, implying the absence of a seasonal bias in brGDGTs-reconstructed temperatures (
Weijers et al., 2011). Measurements of soils and the lake water column also failed to reveal any seasonal trend in the distribution of brGDGTs (
Cao et al., 2018). However, recent studies have indicated that brGDGTs-reconstructed MAAT on the TP may be strongly influenced by winter temperatures (
Li et al., 2017;
Wang et al., 2021;
Sun et al., 2022). Therefore, there is still no consensus on whether there is a significant bias toward warm (or colder) months, which contributes to the difficulty in interpreting temperature reconstructions for the TP. There are also unanswered questions regarding a seasonal bias of alkenone-based temperature variations (
Zhang et al., 2022). Thus, there is an urgent need to obtain a deeper understanding of the seasonality effects inherent in each proxy, associated with their biosynthesis differences.
Transfer functions between temperature proxies (such as brGDGTs and alkenones) and their living environment are an effective method for reconstructing past climate change. It is essential that the calibration model is chosen carefully since the choice of calibration affects both the trend and absolute values of the reconstructed temperature (
Foster et al., 2016;
Wang et al., 2021;
Sun et al., 2022). For example, Wang et al. (
2021) applied a brGDGTs-derived proxy to MAAT reconstruction at Banggong Co using different regional calibrations (
Tierney et al., 2010;
Sun et al., 2011;
Günther et al., 2014;
Wang et al., 2016) and concluded that the record based on the function of
Günther et al. (2014) demonstrated a different pattern of temperature change compared with the other functions. A similar result was found for Ngamring Co (
Sun et al., 2022). Likewise,
Zhang et al. (2022) used all the brGDGTs-temperature calibrations published in previous studies based on the improved separation technique (
Dang et al., 2018;
Russell et al., 2018;
Feng et al., 2019;
Zhao et al., 2021) to reconstruct temperature changes for core CQ1 during the past 60 years; they found that three regional calibrations resulted in a cooling trend, while another calibration resulted in a warming trend. Therefore, the choice of transfer function may also increase the uncertainty of temperature reconstructions, although it has a more significant effect on the absolute temperature than on the temperature trend. This suggests that a regional rather than a global transfer function is needed, considering the unique climate and environment of the TP.
Chronological errors can also affect the interpretation of past climate and environmental changes on the TP (
Hou et al., 2012;
Hou et al., 2019b;
Wang et al., 2022). Thus, the discrepancies among the available paleotemperature records for the TP during the past 5000 years could result from seasonality effects on the temperature proxies, the length of the freezing season of the lakes, the choice of proxy-temperature calibrations, and chronological errors. Further studies with model simulations are needed to address this issue.
5 Conclusions
We have obtained a brGDGTs-based MAAT record from Xiada Co, with the objective of improving our understanding of the temperature history of the western TP over the past ~5000. This record indicates a cooling trend with an abrupt temperature decrease during ~500–300 cal yr BP, which may be linked to the collapse of the Guge Kingdom in the 17th century. A comparison of various late Holocene paleotemperature temperature records for the TP, including that from Xiada Co, reveals conflicting results. Moreover, even the brGDGTs-based MAAT series for the TP are inconsistent. We suggest that the seasonality of the temperature proxies, the length of the freezing season of the lakes, the choice of proxy-temperature calibration, and chronological errors may all contribute to the divergent patterns of temperature changes observed for the TP. Hence, there is an urgent need to develop an increased number of high-quality quantitative palaeotemperature series for the TP, which have unambiguous seasonality, site-specific proxy-temperature calibrations, and robust dating.
PDF
(27833KB)
