Biogeochemical evidence for freshwater periods during the Last Glacial Maximum recorded in lake sediments from Nam Co, south-central Tibetan Plateau

Improved knowledge of deglaciation processes during the termination of the Last Glacial Maximum on the Tibetan Plateau can provide important information for understanding deglaciations in climate-sensitive high-altitude ecosystems. Little, however, is known about this time interval because most lacustrine sediment records from the Tibetan Plateau are younger than 19,000 years. This study focused on a lake sediment record from Nam Co, south-central Tibetan Plateau, covering the interval from ~23.7 to 20.9 cal ka BP. We analysed the distribution and compound-specific hydrogen isotope composition (δD) of sedimentary n-alkanes, as well as the bulk sediment TOC, TN, δ13Corg and δ15N composition, to infer lake system development. Pronounced changes in environmental conditions between ~21.6 and 21.1 cal ka BP, as well as between 23.1 and 22.5 cal ka BP (Greenland Interstadial 2), were inferred from increased aquatic n-alkane amounts and decreased δDn−C23 values within these time intervals, respectively. Freshwater inputs, which most likely resulted from enhanced glacier melting, caused these changes. Our results suggest that mountain glacier retreat on the Tibetan Plateau started earlier than previously assumed. The required energy for thawing was probably provided by temperature changes caused by reorganization of atmospheric circulation, which has also been recorded in Greenland ice records.


Introduction
Global climate change and associated temperature increase have particularly strong effects on highaltitude freshwater resources, through glacier melting, increased evaporation and loss of water-storage capacity. In this regard, the Tibetan Plateau, with the world's third largest amount of glaciers and numerous lakes, represents a key region of interest, especially because it is the main freshwater source area for millions of people in Asia. Nam Co, the second largest Tibetan lake, which is located close to a large glaciated area on the south-central Tibetan Plateau, has already shown a clear response to current warming (Bolch et al. 2010). Hence, investigating past, large-scale climate changes in the Nam Co region, e.g. the Last Glacial Maximum (LGM), can contribute to the understanding of developmental stages during deglaciation, and thus shed light on environmental changes that the region is experiencing today and under future warming.
Because of the more equatorial position of the Intertropical Convergence Zone during the LGM (Chabangborn et al. 2014;Günther et al. 2015), the Indian Ocean Summer Monsoon (IOSM), which today provides moist air masses to the Tibetan Plateau, was much weaker (Barrows and Juggins 2005). As a consequence, the winter monsoon in the interior of the Tibetan Plateau, which is related to the interplay between the Siberian High and the mid-latitude Westerlies, is assumed to have been stronger because of the weaker thermal contrast between the Asian continent and the ocean (Shi 2002). The resulting dominance of colder and dryer air masses may have caused a decrease in annual precipitation, from 1000 to 200 mm, which forced the expansion of continental glaciers under extreme cold during the LGM, resulting in glacier extent on the Tibetan Plateau 15 times larger than today (Shi 2002). Globally, the LGM was characterized by greater extent of ice sheets (Mix et al. 2001), lower sea levels and air temperatures 6-9°C lower than today (Lambeck et al. 2002;Shi 2002). Although the LGM is generally defined as occurring from 26 to 19 ka BP (Clark et al. 2009), the related maximum extent of glaciated areas reveals large spatial and temporal differences (Hughes et al. 2013). The Greenland ice core records, i.e. the most investigated climate archives, show evidence for rapid climate events within the LGM (Lowe et al. 2008). Deglaciation on the Tibetan Plateau, especially in the Nam Co region, is thought to have occurred between 18 and 16 ka BP, similar to the timing of deglaciation in the Southern Hemisphere and thus later than in other regions (Clark et al. 2009). The precise timing of deglaciation in this area, as well as local environmental changes during this period, are still a matter of debate (Clark et al. 2009;Hou et al. 2012).
Although the distribution of n-alkanes and their stable hydrogen isotope composition (dD) have been used successfully to reconstruct Holocene environmental and climatic changes (e.g. humidity and temperature) at Nam Co (Xia et al. 2008;Mügler et al. 2010), such analyses have not yet been applied to LGM sediment records. Sediment n-alkane distribution reflects sources of organic matter input, and the nalkane isotopic composition reflects the signal of source water, which can thus be used to distinguish between meltwater and precipitation water (Aichner et al. 2010b;Günther et al. 2013;Hu et al. 2014).
In this study, analysis of the distribution and dD composition of sedimentary n-alkanes, as well as bulk sediment geochemical analyses (d 13 C org , d 15 N, TOC, TN), were applied to sediments of the late LGM (23.7-20.9 cal ka BP) from Nam Co. These investigations aimed to characterize environmental changes during this time interval at Nam Co and improve our understanding of climate and environmental changes on the broader Tibetan Plateau. Our goal was to improve understanding of regional climate characteristics during the transition from full glacial to deglaciation conditions in this climatically sensitive region. Specifically, we focused on (1) inferring lake development and changes in climate and productivity during this period, and (2) clarifying whether continuous ice cover persisted during this interval and, if not, which processes provided thawing energy to the lake system.

Study area
Lake Nam Co is located in the south-central part of the Tibetan Plateau (30°30 0 -30°56 0 N, 90°16 0 -91°03 0 E) at an elevation of 4722 m a.s.l. (Lin et al. 2008;Schütt et al. 2010) (Fig. 1). The lake is approximately 70 9 30 km, with a surface area *2015 km 2 in 2004 ), a maximum water depth of 99 m ) and a catchment area of about 10,600 km 2 (Wang et al. 2011). Recent climate at Nam Co can be characterized as semi-arid to semihumid, continental , with a mean annual evaporation of 790 mm for the lake surface and 320 mm for the catchment (Li et al. 2008). Most local precipitation is delivered by the IOSM during summer, whereas during winter, cold and dry air masses of the Westerlies and the Siberian High prevail (Chen et al. 2012). Mean annual precipitation varies between 300 and 500 mm ) and the mean annual air temperature is -1°C . Because the Nam Co basin is closed , water balance is controlled mainly by evaporation, precipitation and inflow, but is also influenced, negatively, by seepage of lake water into groundwater (Zhou et al. 2013). Positive water balance today is caused by enhanced input of glacial meltwater from the Nyainqentanglha Mountain Range south of Nam Co, which is a consequence of recently rising temperature (Liu et al. 2010). Vegetation within the catchment is adapted to high-altitude climate conditions and is dominated by alpine meadow and steppe grasses such as Kobresia sp., Stipa sp., Artemisia sp., Oxytropis sp., Morina sp. and Carex sp. .

Field sampling
In 2008, a 10.4-m-long composite sediment core (NC 08/01) was recovered from the south-eastern basin of Nam Co (30°44 0 14.7 00 N, 90°47 0 25.2 00 E), in 93 m of water ( Fig. 1), using a piston corer with a diameter of 90 mm (UWITEC, Mondsee, Austria). This study focused on the lowermost part of the sediment core, 10.38-6.60 m composite depth. This core section was sampled at 5-cm intervals, with *12 g dry sediment for each sample. Additionally, surface sediment samples were taken in 2009 using an HTH gravity corer with a 6.6-cm internal diameter (Pylonex, Luneå, Sweden). The uppermost 1 cm was sampled from these cores. All samples were freeze-dried (P25, Piatkowski, Munich, Germany) and homogenized for further analyses.
Terrestrial plants (e.g. Carex, Kobresia, Stipa, Leontopodium, Potentilla and Dasiphora parvifolia) were collected from the eastern/southern Nam Co catchment and submerged macrophytes (Potamogeton) were taken from 2 to 5 m water depth in the eastern Nam Co basin in 2009. Each plant sample consisted of several individuals. In addition, dung from different herbivores was collected, representing the digested residue of plant matter. Catchment soil samples were taken from two depth intervals (0-5 and 5-10 cm) and sieved with 2-mm mesh. All soil, plant and dung samples were air-dried in the field and transported in paper bags. Particulate organic matter (POM) was filtered from a lake water depth of 15 m using a filtration system with a glass fibre filter of 1 lm pore size to trap fine particles, and pre-filters of 50 and 10 lm to collect coarser particles.
Chronology of Nam Co sediment core NC 08/01 The chronology of the complete NC 08/01 sediment core is based on 24 AMS 14 C ages (Beta Analytic, Miami, USA) obtained from bulk sediment samples (Doberschütz et al. 2014;Kasper et al. 2015 (Kasper et al. 2015). A reservoir effect of 1420 14 C years was inferred from the age of the core sediment-water interface and was subtracted from all radiocarbon ages prior to calibration (Doberschütz et al. 2014). The interval investigated in this study covers a time span of 2.8 ka, between 23.7 and 20.9 cal ka BP. The mean sediment accumulation rate (SAR) in this section of the sediment record is 1.7 mm a -1 .
Measurement of sediment d 13 C org , d 15 N, TOC and TN For geochemical analyses, homogenized aliquots of dry sediment were weighed in tin capsules (*4 mg for C, *80 mg for N). Inorganic carbon was removed by adding H 2 SO 3 stepwise until 120 ll was added (Steinbeiss et al. 2008), prior to measuring the content of total organic carbon (TOC) and the stable carbon isotope composition (d 13 C org ) of the organic fraction. TOC and d 13 C org , as well as the total nitrogen content (TN) and the stable nitrogen isotope composition (d 15 N), were determined in separate runs with a coupled EA-IRMS system (EA: NA 1100 CN-EA, CE Instruments, Milan, Italy; IRMS: Delta C with Delta Plus ion source, Finnigan MAT, Bremen, Germany). Each sample was measured in duplicate to verify sample homogeneity (analytical precision: d 13 C org = ±0.12 %; d 15 N = ±0.12 %). Working standards were acetanilide (d 13 C = -29.8 %, d 15 N = -1.5 %) and caffeine (d 13 C = -40.2 %, d 15 N = -15.5 %), calibrated against international standards IAEA-N1 and NBS-22. The d 13 C and d 15 N compositions are reported in the conventional dnotation relative to the Vienna Pee Dee Belemnite (V-PDB) and atmospheric (AIR) nitrogen standards (Werner and Brand 2001).

Analysis of sedimentary and plant n-alkanes
Lipid extractions were performed on 6 g of dry core sediment and 2 g of dry surface sediment, plant material (all plant parts were used), dung and soil, using an accelerated solvent extractor (ASE 200, DIONEX Corp., Sunnyvale, USA). The ASE was operated at 100°C and 138 bar for 15 min in 2 cycles using a 9:1 (v/v) solvent mixture of dichloromethane and methanol. HCl-activated copper (15 % HCl) was added to remove elemental sulphur. The total lipid extract was then separated using solid phase extraction on silica gel (0.040-0.063 mm mesh; Merck, Darmstadt, Germany) according to Sachse et al. (2006). The n-alkane (aliphatic) fraction was eluted with 80 ml nhexane. Fig. 2 Chronology of Nam Co sediment core NC 08/01 for the investigated interval, as established from the calibrated median ages of five radiocarbon dates on bulk sediment samples (Kasper et al. 2015). Error bars represent the 2r uncertainty range Identification and quantification of n-alkanes from the sediment core samples were accomplished using a GC-FID system (TRACE GC 2000, CE Instruments, Thermo Quest, Rodano, Italy) with a DB-1MS column (30 m length, 0.25 mm ID, 0.25 lm film thickness; Agilent Technologies, Santa Clara, USA) by comparing peak areas and retention times with an external nalkane standard mixture (n-C 10 to n-C 34 ). The PTVinjector was operated in splitless mode with an initial temperature of 45°C for 0.1 min and then heated to 300°C at 14.5°C s -1 for 3 min. The temperature of the GC oven was held at 90°C for 1 min, raised to 300°C at 10°C min -1 , kept constant for 9 min and finally increased at 30°C min -1 to 335°C and held there for 3 min. Helium was used as a carrier gas at a constant flow of 2 ml min -1 . The FID was operated at constant temperature of 300°C with gas flows of 40, 45 and 450 ml min -1 for synthetic air, hydrogen and nitrogen, respectively.
The n-alkane concentrations of recent plant, soil, dung and sediment samples were measured in a GC-FID system (HP 6890, Agilent Technologies, Santa Clara, USA) with an Agilent 19091B-115 column (50 m length, 0.32 mm ID, 0.52 lm film thickness, Agilent Technologies, Santa Clara, USA). The injector was operated in split mode with a ratio of 10:1 and a constant temperature of 280°C. The oven programme started at 80°C and was kept constant for 1 min before increasing at 10°C min -1 to 300°C. After 11 min, the oven heated at 30°C min -1 to 330°C and was kept constant for 5 min. Each sample was measured in triplicate. Results are expressed as ng g -1 dry weight.
To discriminate between the input of fresh and digested allochthonous plant material, the Carbon Preference Index (CPI) for n-C 17 to n-C 31 , which expresses the ratio of n-alkanes with an odd number of carbon atoms relative to those with an even number, was calculated according to: The average chain length (ACL) was calculated according to: Cluster analysis of the sedimentary n-alkane distribution was performed with the PASW statistics 18 software, using K-means clustering. Different sources of lacustrine n-alkanes were visualized by principal component analysis (PCA) using SigmaPlot 10 software.

Measurement of dD of n-alkanes
Stable hydrogen isotope ratios (dD) were analysed using a coupled GC-IRMS system. Because of very low n-alkane concentrations (\250 ng g -1 ), 3-4 sediment samples had to be pooled together. According to the age-depth model, this pooling results in mean dD values spreading over a range of *80 years in the younger section to *200 years in the bottom section. The combined samples were injected into a GC system (HP5890 Series II, Agilent Technologies, Böblingen, Germany) equipped with a BP1 column (60 m, 0.32 mm ID, 0.50 lm film thickness; SGE GmbH, Darmstadt, Germany). The injector was operated at a constant temperature of 280°C in splitless mode. The oven was maintained for 1 min at 60°C, heated at 10°C min -1 to 300°C and was kept constant for 28.5 min. The final ramp heated at 20°C min -1 to 340°C and was kept constant for 3 min. The column flow was constant at 2 ml min -1 . Five percent of the gas flow was transferred to an ion trap mass spectrometer (GCQ Thermo Electron, San Jose, USA) to identify mass fragments and 95 % went to an isotope ratio mass spectrometer (Delta plus XL, Finnigan MAT, Bremen, Germany). The dD composition was determined three times for each sample and reported in conventional d-notation relative to the Vienna Standard Mean Ocean Water (VSMOW). A general offset calculation was accomplished using H 2 reference gas with known isotopic composition and correcting measured dD values from a standard mixture containing several n-alkanes (n-C 10 to n-C 34 ). Outliers were eliminated using Peirce's criterion. The mean standard deviation for all standard triplicates was 5.2 %, consequently no drift correction was applied. The H 3 ? factor was determined on a daily basis and stayed constant within the analytical error at 11.44 (SD = 0.52; n = 9) during the measurement period. J Paleolimnol (2016) 55:67-82 71 Polissar et al. (2009) pointed out that for lowconcentration Tibetan samples, additional concentration-dependent corrections are necessary. We calibrated our GC-IRMS system using standard mixtures of five concentrations (5-200 ng ll -1 ). The dD values increased logarithmically (n-C 23-31 : R 2 = 0.91 to 0.99) below a concentration of 100 ng ll -1 . As the majority of samples were below this limit, a concentration-dependent correction was applied Wang and Sessions 2008). This correction decreased the dD values of small peaks between 2.3 and 4.3 %. However, the correction did not change the significance of differences between the data points in the time series.
In the lower part of the investigated core Section (10.38-8.75 m, 23.7-22.0 cal ka BP) d 13 C org and d 15 N show no clear trends, with mean values of -26.4 ± 0.7 % and 8 ± 0.4 %, respectively ( Fig. 3a, b). Within the upper unit (8.75-6.60 m, 22.0-20.9 cal ka BP) d 13 C org values increase from -27.5 to -24 % and d 15 N values reveal a decreasing trend, from 8.6 to 6.6 %. The two variables are significantly negatively correlated (r = -0.63, p \ 0.01). The mean content and variability of TOC (0.42 ± 0.05 %) are much higher than those for TN (0.05 ± 0.003 %). Hence, the TOC/TN ratio (7.9 ± 0.8) is mainly influenced by changes in TOC content (Fig. 3c, d, e). The lower unit is characterized by a mean TOC/TN ratio of 7.3 ± 0.6, whereas the upper part of the investigated section shows a higher mean value (8.4 ± 0.7) because of increasing TOC content. Between 22.4 and 22.0 cal ka BP, TOC and TN both reach lowest values, and the mean TOC/TN ratio is relatively low (7 ± 0.5).
Distribution and concentration of n-alkanes Catchment (soils, terrestrial and aquatic plants, dung, POM) and surface sediment samples The composition of n-alkanes in terrestrial plant samples, but also in dung as decomposition products of terrestrial plants and soil biomass, reveals an increasing predominance of long-chain n-alkanes n-C 29/31 (ACL = 29, 29.8 and 29.9, respectively), with a clear odd/even preference (ESM Fig. 2). The odd/even preference is also visible in the aquatic plants, although mid-chain n-alkanes n-C 23/25 (ACL = 23.7) are dominant. This bimodal distribution of longand mid-chain n-alkanes within plant biomass is also reflected in the Nam Co surface sediments (ACL = 25.9), showing the generally high contribution of terrestrial plant material to lacustrine organic matter. POM within the lake water is characterized by an n-alkane distribution between n-C 15 and n-C 25 (ACL = 21.1), a low CPI (1.5) and therefore no odd/ even predominance.
A principal component analysis (PCA) of the relative n-alkane distribution from sediments of core NC 08/01, as well as surface sediments, terrestrial and aquatic plants, dung, soil and POM, reveals the different sources of lacustrine organic matter (ESM Fig. 3). The n-alkane n-C 29 has a strong positive loading on Axis 1 (57.6 %), whereas short-to mid-chain n-alkanes n-C 16-25 have negative loadings. The terrestrial signal consists of terrestrial plants, dung and soil, with the former controlled mainly by long-chain n-alkanes n-C 29 and soils more strongly controlled by n-C 31 . Generally, they are clearly distinguished from the aquatic signal, consisting of aquatic plants, surface sediments and POM.
Using the n-alkane distribution, the core was separated into three clusters using cluster analysis (ESM Fig. 4). Sediment samples in cluster I (microbial) have no odd/even predominance of n-alkane chain length and thus a very low CPI (1.4). This cluster is very characteristic for Periods 2 and 5. Sediments in clusters II and III are dominated by odd mid-and longchain n-alkanes. Whereas samples in cluster II (aquatic) show high abundances of the mid-chain nalkanes n-C 23 and n-C 25 , dominance in cluster III (terrestrial) is shifted toward odd carbon numbers Cn-C 23 , particularly toward n-C 31 . The CPIs for clusters II and III are 2.3 and 3.3, respectively. The two clusters show no difference in the abundance of short-chain nalkanes Bn-C 22 . Cluster II is characterized by decreasing abundance with increasing n-alkane chain length [n-C 22 , whereas cluster III reveals a shift towards an increasing trend of n-alkanes [n-C 29 .
dD values of n-alkanes in lake sediments dD data of the most abundant aquatic n-alkane, n-C 23 (dD n-C23 ) in the Nam Co LGM sediments, enable division of the profile into three sections, with two recording a mean value of -156 ± 48 %, i.e. in the bottom and the upper part (Fig. 3j). In the upper part, two excursions appear, with greater values at *21.4 and 21 cal ka BP. Nevertheless, such single events should not be over-interpreted, given measurement uncertainties. The other section differs significantly (p = 0.00, Tukey HSD) from those above and below, with a low mean value of -376 ± 49 % between 23.1 and 22.4 cal ka BP. A moderate decrease is maintained in the terrestrial n-C 29 above 22 cal ka BP (-200 to -350 %), prior to rapid enrichment in the youngest part of the core (Fig. 3k). Because of low biomarker concentrations, continuous interpretable dD records could not be derived for n-C 29 for a period of 960 years, from *23 to 22.2 cal ka BP. A nearly Fig. 3 Climate proxy records for sediment core NC 08/01 between 23.7 and 20.9 cal ka BP. a d 13 C org of bulk sediment, b d 15 N of bulk sediment, c TOC content of bulk sediment, d TN content of bulk sediment, e TOC/TN ratio, f amount of oddnumbered n-alkanes (n-C 17-31 ) and even-numbered n-alkanes (n-C 16-32 ), g calculated Carbon Preference Index (CPI) for nalkanes n-C 17-31 (solid line) and alkane cluster I-III (blue dots) , h amount of n-alkane n-C 23, i amount of n-alkane n-C 29 (solid black line) and n-C 31 (dashed blue line), j dD of n-C 23 and k dD of n-C 29 , where grey dots represent samples below 50 ng ll -1 and blue dots samples above 50 ng ll -1 , l mean sediment accumulation rate (Kasper et al. 2015). Error bars indicate the standard deviation. (Color figure online) 300-year-long data gap occurred for n-C 23 between 22.47 and 22.17 cal ka BP.

Identification of lake periods
The investigated core interval can be divided into seven periods (1-7) according to variations in the bulk geochemical data, n-alkane distribution and dD values (Fig. 3). Period 1 (23.7-23.1 cal ka BP) shows high fluctuations of odd-and even-numbered n-alkanes, while the CPI decreases from 4.7 to 3.1. Stable hydrogen isotope values of aquatic alkanes n-C 23 remain constant (-164 ± 6 %), while n-C 29 increases slightly. Period 2 (23.1-22.5 cal ka BP) is characterized by strongly depleted dD n-C23 values (to -450 %) and the first appearance of alkane cluster I. Period 3 (22.5-22.0 cal ka BP) is characterized by relatively stable and low concentrations of n-alkanes and mean CPI values of 2.7 ± 0.5. Additionally, d 15 N values decrease for the first time. Period 4 (22.0-21.6 cal ka BP) reveals the lowest concentrations of n-C 23 and C 29 . The CPI reaches its maximum at 21.7 cal ka BP because of dominance of odd-numbered n-alkanes, while evennumbered n-alkanes reach minimum concentrations. From this period onward dD n-C29 values generally decrease, while dD n-C23 starts to fluctuate around its mean value, as it did in Period 1. Furthermore, d 15 N values start to decrease, while d 13 C values increase slightly. During Period 5 (21.6-21.4 cal ka BP), the amount of even-numbered n-alkanes increases rapidly, leading to the lowest CPI values. Period 6 (21.4-21.1 cal ka BP) displays the highest concentrations of odd-numbered n-alkanes, especially n-C 23 . Until the end of Period 7 (21.1-20.9 cal ka BP), the CPI ranges around 2.4, while the amount of n-alkanes decreases slightly. Additionally, the terrestrial dD n-C29 signal shows a shift back to enriched values. Cluster II, representing the aquatic signal, mainly dominates during Periods 3, 4, 6 and 7, whereas cluster III, reflecting the terrestrial signal, stands out during Periods 1-4, and cluster I (microbial signal) dominates during Period 5.
Discussion d 13 C org , d 15 N, TOC and TN d 13 C org values of terrestrial C 3 plants typically range around -28 %, whereas those of C 4 plants are clearly higher, *-14 % (Oleary 1981). Hence, relatively low d 13 C org values of organic matter in the Nam Co sediment samples (-26 ± 1 %) may reflect a predominance of C 3 plant material during the LGM. Similar values (-22 to -26 %) are documented throughout the Holocene at Nam Co , indicating no major change in plant source material since the LGM. Recent palynological and botanical studies also defined this high-alpine region as C 3 -dominated, with Kobresia schoenoides, K. pygmaea, Morina, Oxytropis, Stipa, Artemisia and Poa Herrmann et al. 2010). However, as lacustrine algae and aquatic macrophytes (emergent, submerged) also use the C 3 pathway, with d 13 C org values ranging between -26 and -6 % (Aichner et al. 2010b), d 13 C org alone should not be used to distinguish between terrestrial and aquatic C 3 plants. A combined interpretation with the TOC/TN ratio, which is independent of the photosynthetic pathway, can generally be used to better assess the source of organic matter (Meyers and Lallier-Verges 1999). Some considerations, however, reduce the potential benefit of these proxies for detailed paleoenvironmental reconstruction: (1) because of rather constant and low TN values during the LGM, the TOC/TN ratio is mainly driven by TOC shifts, which diminish the applicability of TOC/TN ratio for these sediments, (2) d 13 C org and TOC/TN ratios of aquatic macrophytes from Nam Co and other Tibetan lakes (Aichner et al. 2010b) reveal a broader range and higher values, making it hard to detect terrestrial signals if the macrophyte contribution is high. Nevertheless, the influence of aquatic material in Nam Co is (a) highly diverse under high abundance of these sources and (b) confirmed by the n-alkane signature. As the LGM sediments of Nam Co are largely characterized by TOC/TN ratios \10, sedimentary organic matter was apparently produced mainly by algae and macrophytes, but in certain periods also by terrestrial material like soil, as suggested by the alkane distribution. The decline in d 13 C org values at 22.0 cal ka BP potentially reflects a response to lower salinity from increased precipitation (see dD n-C29 discussion), whereas the following trend back to higher d 13 C org values was likely caused by increased productivity of submerged macrophytes (Aichner et al. 2010c) as a consequence of higher nutrient supply from glacial meltwater and precipitation input (Slemmons et al. 2013) (see discussion of n-C 23 amount).
Further paleoenvironmental information can be derived from trends in the stable nitrogen isotopic composition as an indicator of ecosystem regeneration (Rothe and Gleixner 2000). Because the mean d 15 N value of 8 ± 0.4 % in the oldest Nam Co sediments, until 22.5 cal ka BP (Periods 1 and 2), is higher than values in recent organic material, this may indicate a relatively unproductive ecosystem that used mainly old, recycled nitrogen. The d 15 N values in subsequent periods, however, show a decreasing trend, suggesting a more abundant nitrogen supply by nitrogen fixation, as the lake system started to regenerate under the influence of both terrestrial and aquatic organic matter. Furthermore, the amount of TN reflects the trophic status of the aquatic system. Lower temperatures can decrease the decomposition rate and therefore reduce the deposition of TN within the system. At Nam Co the amount of TN is stable, with a mean value of 0.05 ± 0.003 %, i.e. very low variability compared to the Holocene (*0.2 %) (Doberschütz et al. 2014), indicating general reduced productivity and decomposition rates and therefore low deposition of nitrogen in organic matter during the LGM.

Sources and composition of sedimentary n-alkanes
In general, the distribution and total amount of nalkanes in lake sediments is an indicator of productivity and/or the level of degradation in a lake system or its catchment, enabling the determination of the relative contributions of autochthonous and allochthonous organic matter (Eglinton and Hamilton 1967;Gelpi et al. 1970;Grimalt and Albaiges 1987;Ficken et al. 2000). These measures can be used to reconstruct relative temperature and precipitation (Lin et al. 2008), although large lakes like Nam Co are assumed to react rather slowly to climate change. The main factors that influence the productivity of aquatic plants are temperature, sunlight intensity, nutrient supply and the availability of dissolved inorganic carbon (Lin et al. 2008). The dominant aquatic n-alkane n-C 23 in recent samples, as well as within the older sediments of Nam Co, can be used as a proxy for lake water conditions (Günther et al. 2013). The total amount of n-alkanes in the Nam Co sediments between 23.7 and 20.9 cal ka BP is lower (900 ± 360 ng g -1 ) than in surface sediments (3-28 lg g -1 ) and during the late Holocene (7-28 lg g -1 ) (Günther et al. 2011), indicating lower biomass production because of colder and dryer climate conditions during the LGM. Despite low biomass production, the distribution of short-, mid-and long-chain even-and odd-numbered nalkanes enables one to distinguish changes in vegetation type and source.
To identify the sources of short-, mid-and longchain n-alkanes, investigations of different organic materials from the lake and its catchment are necessary. As shown by Eglinton and Hamilton (1967), terrestrial vascular plants preferentially synthesize odd-numbered long-chain n-alkanes n-C 27-33 , which is confirmed by the n-alkane composition of plant samples from the Nam Co catchment (ESM Fig. 2). Although n-C 29 has previously been interpreted as an indicator for tree vegetation (Schwark et al. 2002), it probably represents herbaceous terrestrial vegetation at Nam Co because it is found mainly in alpine meadows and steppe grasses that dominate the local, treeless vegetation. A distinction between the signal of woody and herbaceous plants is not possible because of mixing of different organic matter sources within the lake sediment (Rao et al. 2011). On the other hand, n-C 29 was also found in higher proportions in emergent macrophyte species Hippuris and lower abundances in submerged plants (Potamogeton, Batrachium), with more affinity to shallow lakes that experience dryer conditions (Ficken et al. 2000;Aichner et al. 2010a).
After grazing and subsequent digestion by herbivores, the plant n-alkane distribution is mainly unaffected with respect to dominance of n-C 29 and n-C 31 . That is, following the nutrition cycle, digested material is retained in the soil where the same n-alkane signature prevails. The input of terrestrial material into the sediment is mainly influenced by the degree of grazing, which entails mobilization of plant material, as well as the strength of aeolian and fluvial transport. Zhao et al. (2003) characterized dust input as the major mechanism transporting terrestrial n-alkanes into a lake system. This aeolian transport is assumed to have been stronger in glacial times than during warm and humid periods. In contrast, mid-chain-length, oddnumbered n-alkanes n-C 21-25 are mainly produced by submerged and floating aquatic plants (Ficken et al. 2000;Aichner et al. 2010a), which is also confirmed by analyses of aquatic plants from Nam Co (ESM Fig. 2). Gao et al. (2011) showed that up to 97 % of the mid-chain-length n-alkane C 23 in sediments is derived from floating and submerged macrophytes. Furthermore, photosynthetic bacteria and algae synthesize mainly short-chain-length, odd-numbered n-alkanes n-C 15-19 , with a dominance at n-C 17 (Gelpi et al. 1970). In addition, some microorganisms produce even-numbered n-alkanes n-C 12-22 (Grimalt and Albaiges 1987), consistent with analyses of POM from Nam Co (ESM Fig. 2) and reflecting a mixture of mid-chain-length n-alkanes from aquatic plants and short-chain-length n-alkanes n-C 17-21 derived from algae and/or microbes.
With the incorporation of allochthonous organic material from the catchment into lake sediments, the n-alkane composition can be altered in two directions. Whereas fresh, undegraded plant material is characterized by a high odd/even preference of n-alkanes and consequently high CPI values (cluster III, ESM Fig. 4), organic matter degraded by soil microorganisms reveals a very low odd/even preference and no chain-length preference (Zech et al. 2011). As a consequence, high input of microbially degraded soil organic matter decreases the sediment CPI (cluster I), whereas intermediate CPI values (cluster II) are characteristic of mainly aquatic-plant-derived organic matter. In general, the sediment organic matter composition of the investigated core section is more strongly influenced by short-to mid-chain n-alkanes compared to long-chain n-alkanes, resulting in closer similarity to the aquatic and POM signal compared to the terrestrial signal (ESM Fig. 3). Regarding the uncertainty caused by mixing processes from heterogenous organic matter sources, additional information about past changes in the ecosystem can be gleaned from the compound-specific stable hydrogen composition.

Hydrogen source for sedimentary n-alkanes
The dD composition of n-alkanes from Tibetan lake sediments has recently been shown to provide information about biological sources (Duan et al. 2011) and past environmental and hydrological changes Günther et al. 2011). This is related to the fact that n-alkane-synthesizing organisms utilize the ambient water as the primary hydrogen source. Hence, aquatic organisms and macrophytes use the lake water, whereas terrestrial plants mainly utilize meteoric water. Previous studies revealed that the isotopic composition of the lake water, which is reflected by the dD of the aquatic mid-chain n-alkane n-C 23 , is modified by evaporation (Günther et al. 2013), whereas the isotopic composition of the precipitation/inflow, reflected by the dD of the terrestrial longchain n-alkane n-C 29, is modified by meteorological conditions like evapotranspiration and relative moisture Xia et al. 2008;Hu et al. 2014). As a consequence, dD n-C23 and dD n-C29 can be used as proxies for aquatic and terrestrial hydrological changes, as well as for investigating lake-level changes . Furthermore, comparison of the terrestrial and aquatic dD values can be used to distinguish between large-scale precipitation changes and inflow events. In the case of Nam Co, it has been shown that depleted dD n-C29 values are assumed to reflect the greater influence of monsoonal rainfall, whereas enriched dD n-C29 values reflect the increased influence of continental moisture sources (e.g. Westerlies) or convective precipitation after evaporation from the lake itself . The trend to depleted dD n-C29 values starting at *22 cal ka BP could therefore be interpreted as a general change to wetter conditions resulting in decreased evapotranspiration. In contrast, depleted dD n-C23 values may reflect the input of isotopically depleted water (e.g. meltwater from winter snowfall), whereas enriched dD n-C23 values likely indicate increased evaporation during summer. Considering this, the strong depletion within Period 2 (23.1-22.5 cal ka BP) could point to initial lake filling by meltwater during the LGM, while dD n-C29 data are missing as an indicator for still under-represented terrestrial vegetation. As a consequence, compound-specific dD values of the different n-alkanes can be used to differentiate between freshwater input from meltwater and precipitation. Throughout the LGM part of the Nam Co sediment record, the aquatic n-C 23 is more enriched in deuterium than is terrestrial n-C 29 . This is typical for closed lakes under arid conditions that have a negative water balance ).

Local paleoenvironmental reconstruction
Period 1 (23.7-23.1 cal ka BP) The first period of the investigated sequence is characterized by large fluctuations in the amount of odd-numbered n-alkanes, probably caused by high sensitivity of the small and shallow lake to changes in nutrient supply (Kasper et al. 2015). According to the dominant terrestrial cluster III, lake sediment was almost exclusively influenced by fresh material from macrophytes (emergent, submerged) as well as soil/dung (mainly n-C 31 ). Slight changes in temperature and water supply would have resulted in a larger response of a shallow lake under glacial conditions. The slight increase in TOC and TN, in conjunction with a fairly stable TOC/TN ratio, implies increased decomposition and thus enhanced accumulation of nitrogen within the organic matter caused by periodically warmer conditions. These changes in temperature are also supported by slightly decreased dD n-C29 values, but did not influence the lake water isotopic signal to a great extent, as inferred from relatively stable dD n-C23 values, as Nam Co had not yet begun to behave like a typical lake.
Period 2 (23.1-22.5 cal ka BP) This period shows a pronounced lake response to climate changes. A remarkable drop in dD n-C23 to strongly depleted values suggests high input of meltwater from the neighbouring Nyainqentanglha Mountain Range at an early stage of deglaciation, when Lake Nam Co was still very small and shallow. If winter precipitation had increased, a thick and longlasting snow cover would have suppressed the otherwise sparse terrestrial vegetation, and even meltwater during summer could not have provided enough moisture to support n-C 29/31 biosynthesis. As a consequence, because of the low amount of the terrestrial biomarker, no detailed isotopic information on terrestrial hydrologic conditions could be gathered for this time interval and beyond. The first occurrence of alkane cluster I supports the appearance of periodic, small erosion events limited to meltwater streams. After the lake experienced its first main freshwater pulse, the climate situation in Period 3 shifted to the driest and coldest conditions within the investigated interval. Limited production of organic matter as a result of cold climate conditions is also reflected by low amounts of n-alkanes in the sediments, inhibiting the reliable determination of the dD of aquatic and terrestrial biomarkers for most of this period. The pronounced reduction of terrestrial organic matter input via lowered erosion and/or productivity is inferred from the absence of C 31 n-alkanes. The major input was derived from aquatic organisms, as reflected by cluster II. Persistent low TOC/TN ratios and decreased d 15 N values point to a further dominance of aquatic organic matter, likely from emergent macrophytes. Decreasing amounts of TOC and TN, however, may reflect reduced within-lake productivity and reduced deposition of organic matter may have beeen a consequence of climate cooling.
Period 4 (22.0-21.6 cal ka BP) During Period 4, the trend toward higher TOC content and therefore higher TOC/TN ratios, as well as decreased d 13 C org and d 15 N values, indicates enhanced input of terrestrial organic material. Additionally, lowest amounts of even-numbered n-alkanes and consequent maximum CPI values suggest the input of fresh material. This increased input is probably related to higher rates of precipitation, indicated by the trend toward depleted dD n-C29 values beginning at 22.0 cal ka BP. Changes in the lake water isotopic signal (dD n-C23 ) are not significant, reflecting an attenuated lake response. There is an apparent inconsistency between increasing TOC content and the decreasing amount of n-alkanes. Nevertheless, lowest amounts of n-C 23 and of n-C 29 indicate cold conditions, which should have reduced productivity in the lake and catchment.
Period 5 (21.6-21.4 cal ka BP) Period 5 is characterized by the highest amounts of even-numbered n-alkanes as well as high n-C 31 input, originating from degraded soil and dung organic matter. The large contribution of reworked material lowers the CPI value to *1 (cluster I). Low CPI values are in accordance with decreasing d 15 N values, reflecting a continuously rising supply of terrestrial organic matter to the lake. This likely reflects erosional processes in the catchment, probably triggered by enhanced runoff from increased precipitation and/or melting glaciers. The dD n-C23 values fluctuate around the mean and dD n-C29 values continue to decrease, suggesting enhanced precipitation, which generated erosion events under warmer and wetter conditions starting at 21.6 cal ka BP, in agreement with increased input of microbially degraded soil material. Because terrestrial vegetation was not yet fully developed, unimpeded precipitation eroded soil material, which is evident in increased sediment accumulation rates (Fig. 3l).
Period 6 (21.4-21.0 cal ka BP) Significant environmental change initiated during Period 5 carried on into Period 6, reflected by the rising amount of aquatic n-C 23 . This strong increase in biomass production of mainly aquatic macrophytes (n-C 23 ), likely induced by enhanced water-column mixing and nutrient supply from freshwater input (Slemmons et al. 2013), is supported by rather low dD n-C23 values and very high d 13 C values. Higher dD n-C29 values are indicative of reduced precipitation. Consequently, the low dD n-C23 values are likely related to inflow of glacial meltwater rather than input of isotopically light rainwater. Additionally, input by terrestrial n-C 31 decreased to the point of absence in some samples.
Period 7 (21.0-20.9 cal ka BP) High variability in almost all climate proxies suggests a system with pronounced climate fluctuations during brief Period 7. d 13 C org and TOC reached maximum values whereas d 15 N reached minimum values at 20.9 cal ka BP. This may indicate high within-lake biomass productivity dominating the contribution to sedimentary organic matter, a result of high nutrient availability from additional input of allochthonous organic material. Thus, this period can be characterized by climate conditions similar to those of Period 1. In addition, high dD n-C29 values also suggest a shift back to arid conditions, i.e. decreased amounts of precipitation.
Although the onset of mountain glacier retreat on the Tibetan Plateau is generally believed to have occurred only between 18 and 16 cal ka BP, our results indicate that first retreat events likely occurred by 23 cal ka BP in the Nam Co region. This might be related to increasing temperatures, which most likely affected glaciers in the Nyainqentanglha Mountain Range south of Nam Co. Meltwater runoff was probably confined to river channels, as the input of reworked soil organic matter was rather limited, illustrated by moderate CPI values. Because the catchment vegetation was not fully developed, precipitation influenced sediment accumulation rates to a greater extent during the LGM. Changing climate conditions, i.e. temperature and precipitation fluctuations, are reflected in cyclic variations of n-alkane amounts during this interval, and initiated the gradual formation of a larger lake.

Regional paleoclimate
The Nam Co sediment record can be compared with relevant paleoclimate records like the Greenland ice cores, Hulu Cave stalagmites (east China), Lake Qinghai sediments (northern Tibetan Plateau) and East Arabian Sea sediments, to reveal connections between regional climate evolution on the Tibetan Plateau and changes elsewhere in the northern hemisphere (Fig. 4). The primary driving force for longterm climate variations was the amount of orbitally forced solar irradiation, which influenced global atmospheric circulation. The main driving air mass at Nam Co during the LGM, and also the one responsible for abrupt climate changes on the northern Tibetan Plateau (e.g. Lake Qinghai), was the midlatitude Westerlies wind system, which is directly controlled by Atlantic meridonal overturning circulation .
Changes in atmospheric circulation on centennial to millennial time scales, however, are not explained by solar irradiation changes (Yao et al. 2001). Two such abrupt circulation changes during the LGM caused the cold Heinrich-2 event (H-2) and the subsequent warm Greenland Interstadial 2 (GI-2) (Fig. 4). The cold/dry and dust-rich H-2 event is recorded between 24.3 and 23.5 cal ka BP in the NGRIP, GRIP and GISP2 Greenland ice core records , in Hulu Cave between 24.4 and 23.6 cal ka BP (Wang et al. 2001), in Lake Qinghai sediments between 24.1 and 23.5 cal ka BP ) and in the East Arabian Sea Surface Temperature (SST) record between 24.2 and 23.3 cal ka BP (Mahesh and Banakar 2014). This indicates an intra-hemispheric coupling of climate events in Greenland and Asia, which can be explained by intensified Westerlies, a consequence of a cold and ice-covered North Atlantic Ocean . It is possible that at *23.5 cal ka BP, shallow Nam Co, which was characterized by dramatic stage fluctuations (Period 1), was influenced by the cold and dry H-2 event. However, because general climate conditions at shallow Nam Co during that time were already dry and cold, evidence from compound-specific data do not support the H-2 onset. The subsequent warmer GI-2 is also visible in the other records, starting at 23.5 ± 0.2 cal ka BP. During Period 2, a significant response is indicated by low dD n-C23 values, indicating increased meltwater input from warmer conditions. The impact of the warmer GI-2 event was less severe on the sparse catchment vegetation, perhaps because the thawing season was too short, providing insufficient time for vegetation regeneration.
Previous studies (Yao et al. 1997) suggested that the Tibetan Plateau, and hence the Nam Co region, reacts more sensitively to climate change than does Greenland. A second freshwater period was detected at Nam Co, starting at 22.0 cal ka BP, in the middle of the colder Greenland Stadial 2c event (GS-2c), which is in general agreement with the GISP2 and GRIP records. After reaching a certain threshold of energy in the system, the precipitation amount or temperature affects the input of nutrients and/or melting. Thus, Nam Co went through several developmental stages from glacial conditions to a deglaciation stage. The freshwater periods at Nam Co could have been precursors for the onset or strengthening of the IOSM during the LGM. Abruptly increasing temperatures would have reduced snow cover on the Tibetan Plateau and therefore increased surface temperature and radiative heating. Consequently, if warming of the Indian Ocean was synchronous with the deglaciation of continental ice sheets, the abrupt freshwater event at Nam Co could be a sign of an early deglaciation pulse Fig. 4 Comparison of proxy records from different locations between 25 and 19 cal ka BP. Greenland Stadials (GS-2b, GS-2c) and Greenland Interstadial 2 (GI-2), following Lowe et al. (2008). The Heinrich-2 event (H-2) is marked, following Moreno et al. (2010). a d 18 O from three Greenland ice cores, as in Rasmussen et al. (2008): GRIP (Johnsen et al. 1997), GISP2 (Stuiver and Grootes 2000) and NGRIP (Andersen et al. 2004), b d 18 O of stalagmites from Hulu Cave, eastern China (Wang et al. 2001), c lake Qinghai Westerlies climate index from flux of [25 lm sediment fraction , d sea surface temperature calculated from Mg/Ca ratios of the planktic foraminifer G. sacculifer in the East Arabian Sea (Mahesh and Banakar 2014), e dD n-C23 and dD n-C29 of n-alkanes in sediment core NC 08/01 from Nam Co, south-central Tibet (this study) on the southern Tibetan Plateau, probably forced by weakening of the winter monsoon and a stronger summer monsoon.

Conclusions
The sediment record from Nam Co, south-central Tibetan Plateau, reflects the early stages of deglaciation during the LGM in a climate-sensitive highaltitude region. Evidence for two distinct freshwater periods was found using stable hydrogen isotope composition, n-alkane distribution and concentration, as well as bulk sediment TOC, TN, d 13 C org and d 15 N composition in sediments deposited between 23.7 and 20.9 cal ka BP. Lake system and catchment development, as responses to intra-hemispheric climate events, were described for seven periods at Nam Co. This study also aimed to determine whether Nam Co was continuously ice-covered during the LGM, and, if not, what factors provided thawing energy. Compared to the Holocene, colder and dryer conditions during the LGM strongly suppressed terrestrial vegetation. Periodic erosion events led to highest sediment accumulation rates until growth of vegetation prevented excessive erosion to shallow Nam Co. The majority of organic matter was produced by aquatic organisms, recording the isotopic signal of lake water. Coldest conditions are inferred for the interval between 22.5 and 22.0 cal ka BP. However, even during that period, ice cover was temporarily interrupted by short thaw periods during summer. High amounts of even-numbered n-alkanes indicate increased erosional input of soil material into the lake between 21.6 and 21.4 cal ka BP. More obvious environmental changes occurred after 23.1 cal ka BP, with the influence of the warmer Greenland Interstadial-2 event leading to very low dD n-C23 values, while dD n-C29 values remained rather high. This likely indicates meltwater inflow from the Nyainqentanglha Mountain Range. A second freshwater period, induced by increased precipitation, was detected by changes in n-alkane distribution and depleted terrestrial dD n-C29 values starting at 22.0 cal ka BP. This period is characterized by several deglacial developmental stages in the middle of the colder Greenland Stadial 2c event. On millennial timescales, environmental changes were caused by a slightly intensified IOSM and attenuated Westerlies during winter. Apparently, the first evidence for deglaciation at Nam Co was preceded by at least two earlier glacier-melting phases between 23.1 and 21.1 cal ka BP, within the warmer Greenland Interstadial 2 and colder Greenland Stadial 2c events.