DO RADIOCARBON AGES OF PLANT WAX BIOMARKERS AGREE WITH 14C-TOC/OSL-BASED AGE MODELS IN AN ARID HIGH-ALTITUDE LAKE SYSTEM?

ABSTRACT To elucidate the dynamics of terrestrial leaf waxes in a high-altitude lake system, we performed compound-specific radiocarbon analysis (CSRA) of long-chain n-alkanes in two sediment core sections from Lake Karakul (Pamirs, Tajikistan) and in surface soil samples from the catchment area. We aimed to answer the question whether the n-alkanes are delivered into the lake sediment with substantial delay due to storage in soils, which may cause a potential bias when used as paleoenvironmental proxies. In the surface soils, the CSRA results reveal an age range of n-alkanes from modern to 2278 ± 155 cal BP. In the two sediment core samples, three of the four n-alkane ages fell on the lower ends of the 1σ-uncertainty ranges of modeled ages of the sediments (based on AMS 14C-TOC and OSL dating results). We conclude that sedimentary leaf waxes represent compounds with intermediate turnover time in soils, for example originating from alluvial plains close to the shores. Overall, the results provide evidence that sedimentary leaf wax compounds in this cold and arid setting are potentially older than the conventional age model indicates, but these findings need to be interpreted in context of the generally large uncertainty ranges of such age models.


INTRODUCTION
Sediment cores are commonly used as climate archives, whose proxy-data interpretation relies on age-depth models. In particular, proxies need to represent the time of deposition, but significant lag times potentially occur on the riverine to marine pathways and in large lake catchment areas before a climate signal is deposited into sediments (Eglinton et al. 1997). A proxy group which is especially prone to such lag times is terrestrial organic biomarkers, e.g., long-chain n-alkanes derived from leafs and roots of terrestrial plants, which are frequently used as hydroclimatic indicators (Eglinton and Hamilton 1967;Castañeda and Schouten 2011). This is mainly because of significant residence time in soils due to high recalcitrance of compounds, their transport via rivers, as well as mixing processes, resuspension, and redistribution of material from floodplains from which it was reworked after a long accumulation time (Pearson and Eglinton 2000;Ohkouchi et al. 2003;Smittenberg et al. 2004;Mollenhauer et al. 2005;Mollenhauer and Eglinton 2007;Eglinton and Eglinton 2008;Vonk et al. 2014;Winterfeld et al. 2018;Berg et al. 2020;Bliedtner et al. 2020).
Lacustrine systems (Uchikawa et al. 2008;Douglas et al. 2014;Gierga et al. 2016;Douglas et al. 2018;Aichner et al. 2018;Yamamoto et al. 2020;Freimuth et al. 2021) have been comparably less studied. Also here, ages of plant waxes from catchment soils range from several hundreds to thousands of years at both temperate and warm/humid study sites. Further, significant offsets between biomarker ages and conventional age models have been observed.
No such data so far exist for a cold and arid high altitude lake system. The main purpose of this study was to identify discrepancies between the ages of plant wax-derived long-chain n-alkanes and the conventional age model, which may result in misinterpretation of the biomarker records. The study was performed in Lake Karakul and its catchment, a well-studied lake system in the Pamirs, Tajikistan (Taft et al. 2014;Heinecke et al. 2017aHeinecke et al. , 2017bHeinecke et al. , 2018Mischke et al. 2017;Aichner et al. 2019; Figure 1). The cold and arid conditions are assumed to result in low carbon turnover and biomarker fluxes to sediments, which may result in significantly higher n-alkane ages compared to the bulk sediment. We measured 14 C contents of long-chain n-alkanes in surface soils of the lake catchment and in the sediment cores. Hereby we aimed (1) to determine the ages of soil-derived n-alkanes, as a likely source of sedimentary biomarkers, and (2) to quantify potential offsets between the ages of n-alkanes and sediments, the latter estimated by means of accelerator mass spectrometry (AMS) 14 C dating of total organic carbon (TOC) and optically stimulated luminescence (OSL) dating.

Study Site
Lake Karakul (approx. 39°N and 73°25 0 E) is a closed-basin lake in the eastern Pamirs, Tajikistan, located in an active extension basin within a graben structure (Strecker et al. 1995;Komatsu 2016). While the lake covers 388 km 2 at an altitude of 3915 m a.s.l., its catchment area extends over 4464 km 2 and reaches up to peaks at 6780 m a.s.l. (Figure 1a). The mean annual precipitation (MAP) within the shielded lower basin is <100 mm, but often exceeds 1000 mm at higher altitudes, mainly due to enhanced winter precipitation. Mean January, July and annual temperatures are −18.1, 8.5, and −4.0°C, respectively (Williams and Konovalov 2008).
The vegetation around the lake consists mainly of species from the Cyperaceae family. At wetter sites they form either dense saline meadows dominated by Carex and Kobresia species or saline marshes dominated by Blysmus species, the latter usually located close to the lake shore. Sites where soil moisture is lower are characterized by sparse vegetation with Carex pseudofoetida. At dry sites, species from the Cyperaceae family are replaced by plants adapted to dry and stony soils, such as Krashenninikovia ceratoides, Artemisia pamirica or Ajania tibetica (Safarov 2003;Mętrak et al. 2017Mętrak et al. , 2019. Since the year 2008, several sediment cores spanning the last ca. 4-31 ka BP were obtained from Lake Karakul and investigated by sedimentological, geochemical and biological analyses (Mischke et al. 2010(Mischke et al. , 2017Taft et al. 2014;Heinecke et al. 2017aHeinecke et al. , 2017bHeinecke et al. , 2018Aichner et al. 2019). The obtained results, and paleoshorelines 35 m above the present lake level, dating to 15 ka BP (Komatsu and Tsukamoto 2015), indicate past lake-level changes, which were mainly driven by changes in the seasonality of the precipitation-evaporation balance (Aichner et al. 2019).

Sampling
The collected soil samples were selected to represent the variety of soils in the catchment and hence cover multiple potential source material of the lake sediment. Pictures of sampling spots are provided in the Supplementary Figures S1. Soil samples #1-3 were collected along a stream that runs from the Sarikol Range on the border between Tajikistan and China and enters a small bay at the northeastern part of Karakul Lake ( Figure 1b; Table 1). Sample #1, located farthest from the lake, was taken from an area covered with sparse vegetation typical for slightly moist places. A dominating species there is Carex pseudofoetida characterized by long creeping rhizomes. It is accompanied by Rheum reticulatum and species from genera Potentilla and Kobresia. Samples #2 and #3 were collected closer to the lake, with sample #3 placed within the stream delta. Both locations are covered by dense wetland vegetation (small sedge meadows) dominated by species belonging to Cyperaceae family (e.g., Carex orbicularis, Carex microglochin, C. pseudofoetida, Eleocharis sp.). Soil samples #4 and #5 were collected on the narrow beach south of Karakul village. The beach represents dense saline marshes dominated by Blysmus rufus and, in drier places, by Carex microglochin. All soil samples were collected as blocks of approximately 10 cm width, 10 cm length and 20 cm depth, cut with a serrated knife and packed separately into linen bags after removal of the surficial live vegetation layer and dense roots. Air-dried samples were crushed by hand, mixed thoroughly and sieved (1-mm mesh) to remove coarse matter and fine roots. Sample #6 was collected from a large alluvial plain close to the southern shore of Lake Karakul. Here, the top 0-10 cm of soil above a layer with increased inclusion of stones/gravel, were sampled.
For sediment samples, two sections of the sediment core KK12-1 (Heinecke et al. 2017a(Heinecke et al. , 2017bMischke et al. 2017) were selected ( Figure 2). Based on the concentrations of nC 29 -and nC 31alkanes (Aichner et al. 2019), the needed sediment amount for CSRA was estimated. Hence, one core section from 488-465 cm depth and another one from 695-669 cm depth were chosen, representing the middle Holocene (6792-5760 cal BP) and the Late Glacial (16,877-14,673 cal BP).

Age Model
Details of age modeling of the composite sediment core KK12-comp have been published previously (Mischke et al. 2017; Figure 2a). Briefly, this age model is based on AMS 14 C ages of TOC and OSL dates and was computed with OxCal 4.2, using IntCal13 (Bronk Ramsey and Lee 2013;Reimer et al. 2013). Two parallel cores (KK12-1 and KK12-2) were correlated by means of elemental data derived from XRF scanning at 2-mm resolution. Samples for 14 C dating of TOC were taken from both parallel cores. Resulting 14 C ages were corrected for a lake-reservoir effect of 1368 yr before calibration. The lake-reservoir (LRE) effect was estimated by using average ages determined on material of two living To allow comparison with this published TOC/OSL age model, 14 C ages of biomarkers as listed in Table 2 have been calibrated with OxCal using IntCal13 (Reimer et al. 2013), despite a newer version of IntCal that has been published in the meantime.

Sample Preparation
All sediment and soil samples were freeze-dried, homogenized and solvent extracted (ASE 350, Dionex, USA) with a mixture of dichloromethane and methanol (9:1 v/v) at 100°C and 75 bar for two cycles. Between 35 and 278 g of dry soil and 57 and 88 g of dry sediment were extracted per sample (Table 1). The aliphatic fraction containing n-alkanes was separated with silica gel chromatography (5 cm × 40 mm; pore size 60 Å, 230-400 mesh size, 40-63 μm particle size) with n-hexane. The aliphatic fraction of the sediment samples was further purified by removing unsaturated compounds with AgNO 3 -coated silica gel. Individual n-alkanes were identified and quantified relative to an external standard by gas chromatography (GC) using an Agilent 7890B system equipped with a flame ionization detector (FID).
Individual n-alkane homologs were isolated by repeated injection on a preparative gas chromatography system (prepGC). The prepGC consists of a 7690 Agilent GC (Agilent, USA) equipped with an ultralow bleed fused silica capillary column (30 m, film thickness 0.5 μm, 0.53 mm ID, RTX-1, RESTEK, USA), a CIS 4 injection system (Gerstel, Germany), and coupled to a preparative fraction collector (PFC, Gerstel, Germany). Individual n-alkanes were collected at room temperature in glass traps attached to the PFC and later flushed with 1 ml dichloromethane. Purity and amount of the isolated compounds were determined by GC-FID. Average recovery of target compounds after prepGC was 24% (Table S1).
In order to achieve quantities large enough for reliable 14 C dating of CO 2 using a gas ion source Melchert et al. 2020), some of the individual compounds of the soil samples were combined as given in Table 2: For Soil #1 nC 29 -and nC 31 -alkanes were pooled, nC 29 -and nC 31 -alkanes from soil #2 and #3, as well as from soil #4 and #5 were combined to one sample respectively. For soil #6 and the sediment samples n-alkane homologues were dated individually. The nC 27 -alkane was not considered for CSRA, as it was not possible to chromatographically separate this homologue from another unknown compound via prepGC. The samples were prepared in two batches at different times. While the isolation with prepGC remained the same, the preparation for 14 C analyses differed. The n-alkanes isolated from the soil samples #1, #2/3 and #4/5 were prepared by sealed tube combustion using the gas injection system (GIS) interface (Stolz et al. 2017), as described in Rethemeyer et al. (2019). In contrast, the n-alkanes purified from soil #6 and sediment core A, B were transferred into tin capsules to be analyzed via direct combustion in an elemental analyzer (EA) coupled with the GIS interface ).

Data Correction
In order to correct the results for possible contamination added during the purification procedure and to determine contributions of extraneous C, two standards of known 14 C Since all samples were expected to be of similar size, two replicates of the modern standard and blank were measured along with the samples and the correction was performed as described in BATS (Wacker et al. 2010). Recoveries of the standards after prepGC accounted for ≈60%. The F 14 C values of the squalane and nC 18 scattered by 1‰ and 20%, respectively. The δ 13 C values scattered by 1.4‰ and 0.2‰. The reported errors include the Poisson errors of the AMS measurements of the standards and of the individual samples. However, according to the model of constant contamination as described in Hanke et al. (2017), an additional error is expected that comes from the measurement uncertainties of the sample masses, which are proportionally larger for smaller sample amounts. The most extreme influence is therefore expected for the sample COL6135 (7 μg), which in fact shows a relatively high deviation of the analyzed isotope ratios of about 70%. With the measured blank values, the constant contamination is estimated to be about 2 μg, assuming a F 14 C value of the contaminant of 0.4.

RESULTS
Except for sample #1, which shows larger relative amounts of nC 27 -alkane, all soil samples are characterized by a pronounced dominance of nC 29 -and nC 31 alkanes. The absolute amounts of extracted n-alkanes range from 13 to 231 μg before and from 5 to 23 μg after prepGC, (Table  S1). The F 14 C values for the soil samples, which are based on either pooled or individual compounds, range from 0.718 to 1.024, equivalent to conventional 14 C ages ranging from 2261 ± 105 BP to modern (-187 ± 102 BP) ( Table 2).

Variability of n-alkane Ages in Surface Soils
The mean residence times of soil organic matter is believed to be relatively short, i.e., a few decades (Kögel-Knabner and Amelung 2014; Feng et al. 2016). However, a number of studies identified a significantly older carbon pool, i.e., several hundreds to a few thousands of years old, which may consist of more recalcitrant or physically stabilized compounds (Schöning and Kögel-Knabner 2006;Smittenberg et al. 2006;Trumbore 2009;Douglas et al. 2014Douglas et al. , 2018Van der Voort et al. 2017).
In the Karakul catchment, surface soils were sampled along a transect from sites covered with sparse vegetation not affected by rivers/streams in the catchment (#1), via soils covered with sedge meadows developed on the alluvial plains of rivers and creeks (pooled sample #2/#3 and sample #6), to samples from the saline marshes in the vicinity to the shore of Lake Karakul (pooled sample #4/#5) (Figure 1). The highest calibrated age of all dated compounds (2278 ± 155 cal BP) was measured in sample #1 (Table 2). The n-alkanes in the pooled saline marsh samples #4/#5 exhibit the second oldest age (1785 ± 92 cal BP). Alluvial samples show mixed results, but principally the lowest ages ranging from modern in #6, to 1250 ± 39 cal BP in the pooled samples #2/#3.
Thus, there is a tendency towards higher n-alkane ages in the surface soils with lowest influence of river water. We hypothesize that there are three major reasons behind this observation: (I) different exposure to ground and surface water and hence proneness to surface erosion and translocation of organic matter (OM). Precipitation amounts in the lower parts Karakul basin are low but significantly higher at higher altitudes. This leads to a strong increase of riverine discharge during the spring/early summer snow melt and large alluvial plains, such as around sampling spot #6, are partially flooded. From soils located in such alluviums or closer to rivers, mobilization and removal of OM is probably efficient and faster. In the saline mashes (spots #4/#5) OM might be protected from translocation due to dense vegetation cover, which could explain the relatively old biomarker age in the sample from those locations.
A second major cause may be (II) different production rates of biomass at wet sites compared to dry sites. We suggest that river flooding delivers nutrients to the soil and supports faster plant growth, resulting in larger inputs of fresh OM and hence a younger organic carbon (OC) pool. Regions further away from rivers are most likely areas of lower OM input and slower OC cyling because they are drier and vegetation is more sparse.
Another possible factor of relevance (III) is the degree of inclusion of deeper soil layers. Subrecent and older OM is more likely included in the C pool beneath the uppermost soil layer (Douglas et al. 2014). Therefore, at spot #6, the smaller sampling depth of 0-10 cm could partly explain the youngest ages of the biomarkers, compared to the other sites where sampling depths were 0-20 cm. The saline marshes (spots #4/#5) are characterized by a gray and loamy B-horizon, located under a thin carpet of living roots, i.e., the contribution of sub-recent OM to the sample might be relatively large.
The variable ages of soils in the lake catchment give evidence for varying accumulation and translocation dynamics of organic biomarkers in the Karakul catchment. This is important in context of the question which kind of material is preferentially transported to the lake and deposited in sedimentary sequences. The data suggest that locations under riverine and lake water influence are less prone to accumulation of OM over longer time-scales. Accelerated and increased riverine discharge after snowmelt could facilitate erosion of topsoils at those spots. Therefore they are more likely originators of biomarkers deposited as part of the lake sediments. In contrast, plant waxes from the sparse alpine meadows, which are mainly precipitation fed, are less prone to be transported to the lake.
In addition to surface water, eolian transport cannot be ruled out as an additional factor for translocation of leaf waxes to the lake. Long range transport has been shown to be an important source of dust in Central Asian mountains (Wu et al. 2009(Wu et al. , 2015Mętrak et al. 2019). It is unclear whether this is relevant for our study area due to the shielded nature of the Karakul basin. Locally, strong winds connected to the mountain-lake system, could at least facilitate short range transport of leaf waxes-either directly from the leaf surfaces or indirectly via dust particles derived from dry soils.

Plant Wax Biomarkers 1583
Age Discrepancies of n-alkanes and Sediments The radiocarbon ages of the sedimentary nC 29 -and nC 31 -alkanes were compared with the published age model, which is based on TOC/OSL ages, of the respective sediment core (Figure 3). Relatively large sections of the core had to be extracted to obtain sufficient concentrations of the long-chain n-alkanes for CSRA, i.e., 23 cm for sample A and 26 cm for sample B (Figure 3). As a consequence, the respective core sections cover relatively wide age ranges of ca. 1000 years (6792-5760 cal BP) for sample A, and ca. 2200 years (16,877-14,673 cal BP) for sample B. Taking the 1σ-uncertainty intervals into account, these ranges significantly increase to ca. 2000 years (i.e., 7254-5274 cal BP; sample A) and >4100 years (18,036-13,911 cal BP; sample B), respectively. 14 C dating of sediments of lakes on the Tibetan Plateau is often complicated by a significant LRE (hard water effect), which might change over time (Lee et al. 2011;Hou et al. 2012;Zhang et al. 2012;Mischke et al. 2013;Lei et al. 2014;Haberzettl et al. 2015;Xu et al. 2021). Hence, if dating of terrestrial plant remains is not possible, radiocarbon dating of bulk TOC might generate wrong age estimates. In case of the sediment core KK12-comp from Lake Karakul, it was attempted to achieve better age control by including OSL ages. In six of total ten cases, OSL dates corresponded well with 14 C dates of TOC (Figure 2a; Mischke et al. 2017), suggesting a relatively constant LRE throughout the core. However, offsets of ca. 5000 years between the two dating methods occurred around 600 cm depth (Figure 3b). Based on a general outlier model run with 5% probability in OxCal (Bronk Ramsey 2009), the respective OSL date of 8600 ± 600 yr BP was considered as outlier and hence omitted in favor of two 14 C-TOC dates from 570 and 600 cm depth (Figure 3b; Mischke et al. 2017).
In three of four cases the calibrated ages of individual n-alkanes fall close to the lowest end of the 1σ-uncertainty interval of the TOC/OSL-age model: For core sample A, the ages of both measured n-alkanes are similar (nC 29 : 7232 ± 183 cal BP; nC 31 : 7134 ± 220 cal BP), taking their age uncertainties into account (Figure 3a). This is despite the fact that on average 2-3‰ higher δ 13 C values of nC 29 in contrast to nC 31 provide evidence for partially different sources of these compounds (Figure 3a; Aichner et al. 2019).
Higher δ 13 C values can be explained by contribution of aquatic macrophytes to the sedimentary nC 29 -alkane pool, which has been frequently observed in high altitude lakes of the Tibetan Plateau (Aichner et al. 2010;Liu et al. 2015) and recently also in a temperate lake system (Andrae et al. 2020). In case of Lake Karakul, macrophytes dominated organic compounds in the middle to late Holocene (Aichner et al. 2019). A partial contribution of macrophytes to the nC 29 pool in core section A is therefore plausible.
The two n-alkanes in sample A are ca. 1000 years older than the average TOC/OSL age (dashed black line in Figure 3a) of the extracted core section. This age difference could indeed be evidence for a delayed sedimentation of terrestrial compounds, due to transport processes and/or "pre-aging" as a relevant factor, indicated by ages of soil n-alkanes. On the other hand, ages of sedimentary n-alkanes are still within the 1σ-uncertainty range of the conventional age model, which excludes a definite statement about such processes.
Similarly, in core sample B, the age of the nC 29 -alkane (17575 ± 573 cal BP) falls close to the lower end but lies still within the 1σ-uncertainty range of the conventional age model. In this case, the nC 29 -alkane is ca. 1900 years older than the average age based on the TOC/OSL-age model (Figure 3b). In contrast, the nC 31 -alkane exhibits a significant age offset (12202 ± 470 cal BP), with an age ca. 5300 years younger than that of nC 29 , i.e., far off the conventional age model. Figure 3 μ and 1σ uncertainty of calibrated radiocarbon ages of nC 29 -and nC 31 -alkanes (green) in the sediment core sections A (top) and B (bottom) in comparison to the age model (black line: median ages; gray line: 1σ uncertainty) based on sedimentary TOC (orange) and on OSL ages (red). To maintain consistency with the published age model (Mischke et al. 2017), all biomarker ages were calibrated with IntCal13 (Reimer et al. 2013). Gray-shaded interval on the depth-axes indicate the core section extracted for CSRA. Arrows towards the x-axes mark the extents of the 1σ-range of the extracted core sections. The vertical dashed black line indicates the average age of the extracted core section based on the published age model (TOC/OSL avg ). Right panels display δ 13 C values of nC 29 and nC 31 alkanes.

Plant Wax Biomarkers 1585
The two possible explanations for the large offset of the nC 31 -alkane age are (I) different sources of the two compounds and overestimation of ages of the conventional age model; or (II) analytical reasons.
Possibility (I) is only plausible with the younger age of nC 31 reflecting the true timing of sedimentation. This would imply significant pre-aging of nC 29 and a conventional age model that provides too old ages for the respective core section. Assuming that the outlier OSL date at 600 cm more precisely reflects the timing of sediment accumulation, the modeled ages in the respective core section would be too old. Similarly, also the modeled ages in the section sampled for CSRA, which are constrained by two 14 C dates of TOC, could be overestimated. However, different sources of nC 29 and nC 31 in the sedimentary pool are unlikely due to the sparse vegetation in the catchment and generally similar δ 13 C values of the two compounds in the respective core section (Figure 3b). An aquatic contribution to nC 29 during the Late Glacial can also be excluded (Aichner et al. 2019). Therefore, if the TOC/OSL-age model is not correct, we would expect similar offsets for both dated n-alkanes. For these reasons we consider possibility (I) as unlikely.
Possibility (II) for the age offset is that the age of nC 31 encompasses larger uncertainty because of analytical reasons. Sample size, i.e., the amount of isolated C after prepGC, is crucial for CSRA (Gierga et al. 2016). The nC 31 isolated from sample B exhibited by far the lowest amount of datable material (7μg C; Table 2). It is therefore the sample most susceptible for contamination with material containing a modern 14 C-signature. In addition, the smaller the sample size, the higher is the potential proportion of contamination, which is unaffected by the applied data calibration and correction strategies. For these reasons, we consider analytical bias due to low amounts of C to be the most likely cause for the age offset between the two n-alkanes in core sample B.
Overall, the mean residence times of n-alkanes, i.e., the offsets of 1000-1900 years between the TOC/OSL-age model and the ages of n-alkanes, are larger than in most previous studies from lacustrine systems (Uchikawa et al. 2008;Douglas et al. 2014Douglas et al. , 2018Gierga et al. 2016;Yamamoto et al. 2020;Freimuth et al. 2021). On the other hand, they are in range of average ages as determined in soils from the Karakul catchment. This speaks for either a mixed origin of compounds, i.e., from multiple sites characterized by variable pre-aging, or for sources mainly from spots with young to intermediate ages of biomarkers, such as the alluvial plains under riverine influence (samples #2/#3 and #6). An explanation for higher offsets in the glacial period (sample B) is higher soil erosion induced by increased summer humidity in the Late Glacial (from ca. 17 to 14 ka BP; Aichner et al. 2019), which could have led to mobilization of older OM that has been stored in soils during the glacial.

CONCLUSIONS
Three of four dated compounds in two sediment core samples are 1000-1900 years older than average sediment ages modeled by means of 14 C-TOC and OSL dates. On the other hand, these plant wax ages still fall within the lowermost margins of the age models 1σ-uncertainty ranges, which do not yet include additional factors such as changes of LRE.
Based on ages of biomarkers extracted from soil samples, we hypothesize that sedimentary biomarkers are most likely originated from spots with short to intermediate turnover time of organic matter. These are for example the alluvial plains close to the southern and eastern shores and comparable spots with relatively short transport paths and times via surface runoff.
Overall, the data suggest that terrestrial n-alkanes are potentially older than the TOC/OSLbased age model indicates. However, in context of the relatively large uncertainties of such age models, the offsets to plant wax ages appear moderate. We therefore conclude that conventional age models can be potentially applied to terrestrial biomarkers, but care must be taken especially when interpreting such proxy data with respect to short-term events.