Palaeoclimate Reconstruction of the Central Gangdise Mountains, Southern Tibetan Plateau, Based on Glacier Modelling

Abstract

Palaeoglacier modelling is an important approach for reconstructing the palaeoclimate. The timing of glaciations in the central part of the Gangdise Mountains has been constrained previously, but the palaeoclimate remains unclear. In this paper, the palaeo-temperature and precipitation of the early marine isotope stage (MIS) 2, the Last Glacial Maximum (LGM), and the early Holocene were reconstructed using coupled mass balance and ice flow models. The results show that a series of temperature changes (ΔT) and precipitation factors (F_p) resulted in optimum palaeoglacial extents. The modelled palaeoglaciers during the early MIS 2, the LGM, and the early Holocene cover areas of ~18.1 km², ~17.4 km², and ~16.3 km², respectively, with ice volumes of ~2.18 km³, ~1.99 km³, and ~1.95 km³, respectively. Previous studies on ice cores, pollen samples, and lake sediments were referenced to narrow the range of palaeo-temperatures and precipitations. The reconstructed temperatures during the early MIS 2, LGM, and early Holocene were constrained to 2.4–2.9 °C, 2.15–3.05 °C, and 0.95–1.5 °C lower than today, respectively. Their precipitation levels were 60–80%, 50–80%, and 100–150% of the present-day level, respectively.

1. Introduction

The Tibetan Plateau (TP) has the highest number of glaciers in the world outside the polar regions [1]. A complete understanding of the glacial history of the TP is of great significance to better understand climate change in this region and the rest of the world [2]. With the development of numerical dating techniques (especially ¹⁰Be exposure dating and optically stimulated luminescence), glacial chronology has been studied in many places of the TP and the timings of glaciations have been constrained [3,4,5]. Such studies have also been conducted in the Gangdise Mountains [2,6,7,8,9], which are assumed to be one of the mountain regions that first reached the cryosphere on the TP [10]. To identify changes of glacial extent, these studies focused on changes of the glacial equilibrium line altitude (ELA), as these changes are related to climate. However, ELA is also affected by topographic factors and cannot provide climatic conditions directly [11,12,13]. In cases where geomorphic evidence is limited, glacier modelling becomes a very effective tool for inferring the palaeoclimate [14,15,16,17]. The past 10 years have witnessed the applications of glacier modelling on the TP, e.g., in Tashkurgan [18], Yingpu valley [19], Kuzigun valley [20], Payuwang valley [21], Yuqiongqu and Barenduo valley in the Samdainkangsang peak [22], Nyaiqentanglha Shan [23], and Quemuqu valley [24] (Figure 1A). These modelling results showed that the reconstructed temperatures and precipitations differ slightly between different places (see Section 5.2 and

Table 1. Modern temperatures, precipitations, sensitivity tests, reconstructed equilibrium line altitudes (ELAs), temperature, and precipitation during the Last Glacial Maximum (LGM) of the study sites on the Tibetan Plateau.

Study Site	Modern MAAT (°C)	Modern MAP (mm)	ΔT Change (°C) As a Result of a DDF Change of ±0.5 mm °C−1 Day−1	Fp Change As a Result of a DDF Change of ±0.5 mm °C−1 Day−1	ΔT Change (°C) As a Result of a Ts Change of ±1 °C	Fp Change As a Result of a Ts Change of ±1 °C	ELA during the LGM (m asl)	ΔELA (m) (Relative to the Modern ELA)	Temperature Decrease (°C)	Precipitation Relative to Modern Value (%)	References
Tashkurgan valley	3	68.9	±0.6	±0.1	±0.5	±0.1	4550–4600	~600	5–8	30–70	Xu et al. [18]
Yingpu valley, Queer Mountains	6.9	623.4					4630	500	4–5.9	40–80	Xu [19]
Kuzigun valley, Lyavirdyr Tag	3.9	78.6	±0.9	±0.25	±0.8	±0.25			5–8	30–70	Xu et al. [20]
Payuwang valley, Nyaiqentanglha Shan	0	282					5410	340	3.3–4.4	30–70	Xu and Glasser [21]
Barenduo valley, Samdainkangsang Peak	2.06	478									Xu et al. [22]
Yuqiongqu valley, Samdainkangsang Peak	2.06	478									Xu et al. [22]
Nyaiqentanglha Shan	2.06	478	±0.65	±0.25	±0.7	±0.25			~3.6	~50	Xu et al. [23]
Quemuqu Valley, Qiongmu Gangri Peak	2.06	479.3	±0.55	±0.4	±0.6	±0.2	5320–5380	350–410	3.1–4.3	30–70	Xu et al. [24]
Gaerqiong Valley, central Gangdise Mountains	−8	544	±0.05	±0.01	±0.1	±0.02	~5816	~124	2.15–3.05	50–80	This study

), which may imply the influence of local topoclimatic factors on glaciers. This highlights that palaeoglacier modelling needs to be applied to different regions (especially with different climates) to identify patterns of palaeoglacial and climatic fluctuations.

This study focuses on the Lopu Kangri area of the central Gangdise Mountains, which is dominated by a weak monsoonal climate. The timing of glaciations of the Lopu Kangri area has been constrained at ≥243.88 ± 25.88 ka, ≥43.09 ± 4.18 ka, 24.19 ± 2.29 ka, 19.78 ± 1.90 ka, 10.62 ± 1.00 ka, 2.75 ± 0.37 ka, 1.80 ± 0.18 ka, 0.32 ± 0.04 ka, and 0.22 ± 0.04 ka using ¹⁰Be exposure dating [2]. However, the palaeoclimate at the time of these glaciations has not been reconstructed. Zhang et al. [2] compared the ELA depressions (ΔELAs, relative to the modern value) during several glaciations on the TP and found that, because of the arid climate of the Gangdise Mountains, ΔELAs of the Gangdise Mountains are smaller than those of other mountains. This trend implies that the nature of glaciations in the Gangdise Mountains had a different characteristic compared to those in other climatic domains. The palaeo-temperatures and precipitations in the Lopu Kangri area of the central Gangdise Mountains were reconstructed during the early marine isotope stage (MIS) 2 (24.19 ± 2.29 ka), the Last Glacial Maximum (LGM, 19.78 ± 1.90 ka), and the early Holocene (10.62 ± 1.00 ka). A mass balance coupled with an ice flow model was used to reconstruct the palaeoclimate during the early MIS 2, the LGM, and the early Holocene. The results were compared with those obtained for other places on the TP to deepen the knowledge on palaeoglacier fluctuations and palaeoclimate variations during the late Quaternary.

2. Study Site

The Gangdise Mountains are located in the south of the TP along a NW–SE axis [11]. These mountains form one of the most prominent tectonic units on the TP, with many mountain ranges rising above 5000 m above sea level (asl). The central-eastern sector of the Gangdise Mountains is dominated by the Indian summer monsoon (ISM) whilst the westernmost sector is dominated by westerlies [26]. The present study focuses on the glaciers in the Gaerqiong Valley, east of Lopu Kangri (29°50′ N, 84°36′ E, 7095 m asl) in the central sector of the Gangdise Mountains (Figure 1B), which contains a main valley and a tributary valley. The main valley has a length of ~5.3 km and a width of ~2.3 km and is occupied by a valley-type glacier [2]. The southern bank of the Gaerqiong Valley is a tributary valley, occupied by another valley-type glacier. The glaciers of the study area cover a total area of 7.7 km², with thicknesses reaching up to 137 m [27]. Twelve frontal moraines (M1–12) can be observed in the Gaerqiong Valley. Using the cosmogenic ¹⁰Be exposure dating method, Zhang et al. [2] constrained nine glacial events in the Gaerqiong Valley and adjacent areas. The moraine that formed during the early Holocene (M8) is located ~4 km away from the mouth of the glacier, ~5–10 m above the bottom of the valley, and has a length of ~4.3 km. The frontal moraine formed during the LGM (M9) is located ~4.6 km downstream of the modern glacier, with a length of ~680 m and an elevation ~5–30 m above the valley floor. M10 (which formed during the early MIS 2) was incised into the southern and northern sections by glacial streams. The southern section is located ~300 m from M9. It has a length of ~320 m and an elevation of ~10 m above the bottom of the valley. The northern section has a length of ~440 m and an elevation of ~20 m above the valley floor. Sub-rounded or sub-angular granite boulders were found on the surfaces of these moraines, and some boulders showed residual lacquer on the surfaces [2]. According to the High Asia Refined Analysis (HAR) v2 dataset (https://www.klima.tu-berlin.de/ (accessed on 10 October 2021)), the mean annual air temperature (MAAT) of the study area is −8 °C, and the mean annual precipitation (MAP) is 544 mm, with most precipitation occurring during summer months [28] (Figure 2).

3. Methods

A mass balance coupled with a two-dimensional (2D) ice flow model was used to simulate the palaeoclimate.

3.1. Model Input

The input data include the digital elevation model (DEM) of the subglacial surface, the monthly mean temperature (°C), and the monthly mean precipitation (mm). The DEM of the subglacial surface was derived from the Shuttle Radar Topography Mission 1 arc-second global elevation data v4.1 (https://earthexplorer.usgs.gov (accessed on 13 October 2021)) after extracting the thicknesses of modern glaciers, which were derived from the dataset of Farinotti et al. [27]. Monthly air temperature and precipitation were derived from the HAR v2 for the period of 1991–2020 (https://www.klima.tu-berlin.de/ (accessed on 10 October 2021)) [28].

3.2. Mass Balance Model

The glacier net mass balance (MB) (mm a⁻¹) was estimated with a positive degree-day method [29,30]:

$$MB={{\displaystyle \sum }}^{}P-{{\displaystyle \sum }}^{}{T}^{+}\times DDF$$

(1)

where P is the mean monthly snowfall (mm), T⁺ is the monthly sum of positive air temperature (°C), and DDF is the degree-day factor (mm °C⁻¹ day⁻¹) [30]. Snowfall is calculated as a fraction of the month’s total precipitation when the air temperature reaches the snowfall threshold (T_s = 1 °C) [16,31]. ELA was regarded as the mean elevation of the boundary of the glacier’s accumulation zone (where MB is above 0) and ablation zone (where MB is below 0).

3.3. Ice Flow Model

The mass transfer between accumulation and ablation zones was calculated based on the ice flow model used by many previous studies [14,15,16], which is expressed as follows:

$$\frac{\partial h}{\partial t}=MB-\frac{\partial q_x}{\partial x}-\frac{\partial q_y}{\partial y}$$

(2)

where h is the ice surface altitude, t is time, and q_x and q_y represent ice flux along the x and y dimensions, respectively, in the horizontal plane. The ice flux between adjacent DEM cells (q_x, q_y) is determined by ice thickness (H) and mean vertical ice flow velocity (u). u comprises ice deformation (u_d) and sliding (u_s), and is calculated using Equation (3):

$$u=u_d+u_s=\frac{2}{5}\left(1-f\right)HA{\tau }^m+fB{\tau }^n$$

(3)

where f is a factor that adjusts the fraction of flow caused by ice deformation and sliding, A and B are coefficients, τ is basal shear stress, which is calculated using Equation (4). In this study, f, A, and B are set to 0.5, 1 × 10⁻⁷ Pa⁻³ a⁻¹, and 1.5 × 10⁻³ m Pa⁻³ a⁻¹, respectively [14]:

$$\tau =\nabla \rho gH$$

(4)

where ρ is ice density and g is gravitational acceleration.

The ice flow model is used to evolve glacier geometry by calculating ice velocity offset from points of known ice thickness. The flux gradient is then used to calculate an updated ice thickness using an explicit forward time step [22].

However, there are some limitations of the model. For example, the input of DEM, temperature, and precipitation produces uncertainties, especially as no weather stations are available in the study area. The parameters of the model also produce uncertainties [19]. The associated uncertainty has been widely discussed in previous studies and this paper does not attempt to address this issue [15,18,32,33].

3.4. Modelling Strategy

The modern glacial extent was first modelled under the current climate with varying DDF values. The DDF value resulting in the glacial extent that matches the observed extent is regarded as the optimum DDF value, which was then used in the palaeoglacier modelling. To model palaeoglaciers, different combinations of temperature change (ΔT, °C) and precipitation factor (F_p) were applied until the modelled palaeoglacial extents matched observed extents. These ΔT/F_p combinations represent the possible climate during glaciations.

The results show that the optimum DDF value is 14.7 mm °C⁻¹ day⁻¹. Under this DDF value, the simulated extent of modern glaciers is in good agreement with the actual situation (Figure 3). The ELA of modern glaciers, simulated under optimal conditions, is ~5940 m asl. DDF and T_s are the main parameters affecting the estimation of the palaeoclimate in this model [18]. To test the sensitivity of DDF values and T_s, multiple runs with varying DDF values (14.2 and 15.2 mm °C⁻¹ day⁻¹) and T_s (0 and 2 °C) were employed. ΔT/F_p combinations were changed until the glacier reached the observed position. When testing the sensitivity of the DDF (T_s), T_s (DDF) was set to 1 °C (14.7 mm °C⁻¹ day⁻¹).

4. Results

4.1. Early MIS 2

The ΔT/F_p combinations required to reproduce a good match to the early MIS 2 moraine location can be identified in Figure 4, with ΔT ranging from −3.4 to −2 °C and F_p ranging from 0.5 to 1.0. Temperature varies greatly under relatively dry conditions, but this trend diminishes under wet conditions (Figure 4). This indicates that under dry conditions, a suitable glacier range could be generated by small temperature changes. Moreover, wetter conditions produced steeper mass balance gradients (with higher rates of accumulation and ablation), while a colder climate produces flatter mass balance gradients. This results in smaller accumulation areas under wet conditions and larger accumulation areas under cold conditions. For early MIS 2, for example, the ΔT/F_p combinations of −2.9 °C/0.6 and −2.4 °C/0.8 produced accumulation-area ratios (AAR) of 0.83 and 0.73, accumulation areas of ~15 and ~13 km², respectively. The modelled glacier during early MIS 2 covers an area of ~18.1 km² and has a volume of ~2.18 km³. It has a maximum thickness of ~275 m and a minimum height with ice of ~5306 m asl (Figure 5A,B). The modelled mean ELA during the early MIS 2 is ~5797 m asl, which is ~143 m lower than the modern value.

4.2. LGM

All combinations that reproduce the glacial extents during the LGM are shown in Figure 4 (ΔT ranges from −3.05 to −1.8 °C, F_p ranges from 0.5 to 1.0). Under dry conditions, ΔT shows a fast change rate with changing F_p, but a slow change rate when F_p has a high level. This trend is consistent with the trend during the early MIS 2 (Figure 4). Similarly, the mass balance gradient and mass accumulation area are similar to the early MIS 2 period. For instance, the ΔT/F_p combination of −2.15 °C/0.8 and −3.05 °C/0.5 produced AAR values of 0.71 and 0.86, accumulation areas of ~12 and ~15 km², respectively. This means that ice accumulates less in warm and wet conditions and more in cold and dry conditions [22], which is in line with the result of Xu et al. [19]. The modelled glacier during LGM covers an area of ~17.4 km², with a volume of ~1.99 km³ (Figure 5C,D), a thickness of up to ~269 m, and altitude of ~5335 m asl in the glacier snout. The modelled glacial ELA during the LGM is ~5816 m asl., which is ~124 m lower than the modern value.

4.3. Early Holocene

The ΔT/F_p combinations (ΔT ranges from −1.5 to −0.65 °C, F_p ranges from 1.0 to 1.9) required to reproduce a good match with the early Holocene moraine position can be identified in Figure 4. With changing F_p, the variation of ΔT values shows a roughly linear trend (Figure 4). Temperature is almost equally sensitive to the development of glaciers under both wet and dry conditions. The modelled glacier during the early Holocene covers an area of ~16.3 km², with a volume of ~1.95 km³ (Figure 5E,F), a maximum thickness of ~277 m, and the glacier snout is located at ~5368 m asl. The large glacier thickness during the early Holocene is likely due to the high precipitation of that period. The ELA of the early Holocene is ~5830 m asl., with an ELA that is ~110 m lower than the modern value.

4.4. Sensitivity Test

If F_p remained unchanged, a variation of ±1 °C in T_s produces a variation of ΔT of ±0.1 °C. For example, when F_p is set to 0.5, the ΔT values increase from −3.15 °C to −2.95 °C when T_s values increase from 0 to 2 °C. If ΔT remains unchanged, a variation of ±1 °C in T_s produces a variation of F_p of ±0.02. For example, at a ΔT level of −2.8 °C, the F_p values decrease from 0.58 to 0.54 when T_s values increase from 0 to 2 °C (Figure 6).

A variation of ±0.5 mm °C⁻¹ day⁻¹ in DDF results in a ±0.05 °C difference in ΔT if precipitation remains unchanged. For example, at a F_p of 0.5, the ΔT values decrease from −3.0 to −3.1 °C when DDF values increase from 14.2 to 15.2 mm °C⁻¹ day⁻¹. Then, a variation of ±0.5 mm °C⁻¹ day⁻¹ in DDF produces a variation of F_p of ±0.01 (if ΔT remains unchanged). For instance, at a ΔT of −2.8 °C, F_p values increase from 0.55 to 0.57 when DDF values increase from 14.2 to 15.2 mm °C⁻¹ day⁻¹ (Figure 6).

5. Discussion

5.1. Sensitivity Analysis

A variation of ±0.5 mm °C⁻¹ day⁻¹ in DDF results in a ±0.05 °C different in ΔT (Figure 6). Taking the LGM as an example, with a DDF variation of ±0.5 mm °C⁻¹ day⁻¹, previous studies [18,20,23,24] in other places of the TP showed that the ΔT varied within a range of ±(0.5–0.9) °C, and F_p varied within a range of ±(0.1–0.4). Both are greater than the ranges of ΔT/F_p variations in the Gangdise Mountains (Table 1). According to Equation (1), the DDF value is negatively affected by the monthly sum of positive air temperature (T⁺), and under a colder climate, the DDF variation has less effect on glacial MB and ice thickness. A variation of ±1 °C in T_s produces a variation of ΔT with only ±0.1 °C in the central Gangdise Mountains (Figure 6). With a T_s variation of ±1 °C, previous studies [18,20,23,24] in other places of the TP showed that ΔT varied at a range of ±(0.5–0.8) °C, and F_p varied at a range of ±(0.1–0.25), which exceed the ranges of ΔT/F_p variations in the Gangdise Mountains (Table 1). One possible explanation for this is that in the central Gangdise Mountains, the monthly temperatures (other than June to September) are far below 0 °C, and a variation of ±1 °C in T_s does not greatly affect the modelled glacial MB and ice thickness. These comparisons in turn imply that variations of DDF and T_s in the Gangdise Mountains have less effect on the modelled results compared to other previous studies, i.e., uncertainties of the modelling caused by uncertainties of DDF and T_s in the Gangdise Mountains are less impactful compared to previous studies.

5.2. Palaeoclimate during the Early MIS 2, LGM, and the Early Holocene

To narrow the ranges of ΔT and F_p, the obtained results were compared to other palaeoclimatic reconstruction results. Marine benthic foraminifera δ¹⁸O records identify the early MIS 2 as a cold period, characterised by a low solar radiation level in the Northern Hemisphere [34,35] (Figure 7). High-resolution lake pollen and grain size records in the Balikun Lake indicate a dry and cold climate during MIS 2 [36]. Shukla et al. [37] found that the temperature during MIS 2 (~25–22 ka) was a ~3 °C lower than the present, and the precipitation was ~30–40% of today’s level in the Dingad Basin of the central Himalaya Mountains. Pollen analysis from the upper part of the RM core (33°57′ N, 102°21′ E, 3401 m asl) in Zoige, northeastern TP, shows that during the early MIS 2, the climate was cold and dry, the annual temperature was 5–6 °C lower than today, and the precipitation was only 60–80% of the present-day level [38]. Based on these results, precipitation during the early MIS 2 in this study area was narrowed to 60–80% of the present-day level, with temperatures 2.4–2.9 °C lower than today.

The δ¹⁸O record of the Guliya ice core places the LGM in a relatively cold period [39] (Figure 7). During the LGM, the global glacier volume reached its maximum and the overall climate was cold and dry [40]. The Asian monsoon index obtained from the sediment of Qinghai Lake indicates that the LGM period was characterised by a cold and dry climate [41]. In addition, several studies that used atmospheric general circulation models and regional climate models inferred that the climate during the LGM was generally colder and drier than today [42,43,44,45]. Ju et al. [45] found that the simulated annual average surface temperature of East Asia was 2–4 °C lower than today. According to the RM core pollen records in the Zoige region, the LGM was found to have had a precipitation of only 60–80% of the present and a temperature of 5–6 °C lower than today [38]. By summarising different data methods in different locations of the TP, Shi et al. [46] estimated that during the LGM, the precipitation was about 30–70% of the modern level and the temperature was about 6–9 °C lower than the modern level. In reference to these results, in this study, precipitation during the LGM was narrowed to 50–80% of the modern level, with the temperature being 2.15–3.05 °C lower than today.

The reconstructed temperature during the LGM in the head of Tashkurgan Valley was 5–8 °C lower than the present-day temperature, and the precipitation was 30–70% of the present [18]. The reconstructed climate of the Yingpu Valley in the eastern TP shows that during the LGM, temperature decreased by 4.0–5.9 °C, with precipitation levels 40–80% of the modern level [19]. The modelling result shows that temperature decreased by 5–8 °C and precipitation levels of 30–70% of modern amounts during the LGM in the Kuzigun valley, Lyavirdyr Tag [20]. Xu and Glasser [21] inferred that the LGM had a 3.3–4.4 °C lower temperature and a 30–70% lower precipitation compared to the present in the Payuwang valley, Nyaiqentanglha Shan. Xu et al. [23] found that the temperature dropped by ~3.6 °C during the LGM and the precipitation was 50% of the current-day level in Nyaiqentanglha Shan. Xu et al. [24] showed that a 3.1–4.3 °C lower temperature and a 30–70% lower precipitation level than today conforms well to the climate of the LGM in the Quemuqu valley. By studying the change of ELA in western China during the LGM, it can be concluded that the summer temperature of the TP generally decreases by ~4 °C and the annual precipitation decreases by 25% [47]. The precipitation level during the LGM as reconstructed in this study is similar to the values mentioned above; however, the temperature decrease reconstructed in this study is smaller than the values reported before. This implies that in the central Gangdise Mountains, the temperature increase since the LGM is relatively low. Given that the sites of the previous studies are located in the marginal regions of the TP (Figure 1), they are affected by the westerlies or ISM more significantly, and thus they have experienced larger temperature variations since the LGM than that of the central Gangdise Mountains.

The average ΔELA since the LGM in western China, as simulated by the high-level resolution climate model, was 393–1440 m, with a ΔELA of <500 m in the central TP [47]. In this study, the simulated ELA of the LGM is ~5816 m asl, which is ~124 m lower than that of modern value. In the Tashkurgun valley, northwest TP, the modelled ELA in the LGM ranges from 4550 to 4600 m asl and ELA is ~600 m lower than the modern value [18]. The ELA during ~16 ± 2 ka was estimated as ~4852 ± 218 m asl, which is ~522 m lower than that of the modern glaciers in the semi-arid Suru Basin, western Himalaya Mountains [25] (Figure 1). The simulated ELA in Yingpu Valley of the eastern TP is ~5130 m asl and ~4630 m asl during the LGM, and the ELA in this area is ~500 m lower than today [19]. Xu and Glasser [21] showed that in the Payuwang valley of western Nyaiqentanglha Shan, the simulated ELA during the LGM is ~5410 m asl, which is ~340 m lower than the present-day value. The simulated ELA during the LGM period in the Quemuqu Valley region of the southern TP is ~350–410 m lower than the ELA of modern glaciers (5730 m asl) [24] (Table 1). Compared with these results, it is found that the ΔELAs in the central-southern part of the TP since the LGM are lower than those of eastern and western areas (Figure 1A and Table 1). The eastern part of the TP is dominated by the ISM, while the western part is dominated by the westerlies, both of which receive more moisture than the central-southern part of the TP. This distribution pattern of moisture may explain the spatial pattern of ΔELA of the TP since the LGM.

The Younger Dryas period (11–10 ka) in Europe and eastern North America (studied using palynological, pollen, and simulations) had a cool and wet environment [48,49]. According to the records of lake sediments on the TP, 11–10 ka was a cold period that reached its peak at 10.7–10.5 ka [50]. Ma et al. [51] and Chen et al. [52] summarised various temperature proxies such as pollen assemblages, ice core δ¹⁸O, and glycerol dialkyl glycerol tetraethers during the Holocene, and identified the early Holocene as a cold period, which was in line with the record of low CO₂ concentration [53]. The early Holocene was characterised by increasingly wet weather events based on pollen data of the southwestern TP [51]. Studies on precipitation variability of the East Asian summer monsoon showed that the effect of the monsoon gradually increased from ~14.7–7.0 ka, and pollen-based precipitation reconstructed from Gonghai Lake during the early Holocene was ~20% higher than that of the present [54]. Shukla et al. [37] found that the temperature during the early Holocene (~8 ka) was ~1.3 °C lower than that at present in the Dingad Basin of the central Himalaya Mountains. Compared to these results, the reconstructed precipitation level during the early Holocene in this region was narrowed to 100–150% of the modern-day level, with a temperature 0.95–1.5 °C lower than today.

6. Conclusions

Understanding the fluctuation of the palaeoglacial extent and palaeoclimate variations during the late Quaternary is of great significance to better understand the climate change of the TP and even the world. This study focused on reconstructing the palaeoclimate during the early MIS 2, LGM, and early Holocene in the Gangdise Mountains of the southern TP using a mass balance model coupled with a 2D ice flow model. The results showed that glacial areas and ice volumes in the Gangdise Mountains during early MIS 2, LGM, and early Holocene were ~18.1 km²/~2.18 km³, ~17.4 km²/~1.99 km³, and ~16.3 km²/~1.95 km³, respectively, with ELAs of ~5797 m asl, ~5816 m asl and ~5830 m asl. In reference to the results of previous studies, the reconstructed temperatures during the early MIS 2, LGM, and early Holocene of the central Gangdise Mountains were narrowed to 2.4–2.9 °C, 2.15–3.05 °C, and 0.95–1.5 °C lower than today, with precipitation levels 60–80%, 50–80%, and 100–150% of today’s levels, respectively. The modelled precipitation during the LGM in the Gangdise Mountains was comparable to the results of previous studies on the TP. However, the reconstructed level of temperature decrease in the Gangdise Mountains was lower than them, implying that the Gangdise Mountains have experienced a lesser temperature increase since the LGM.