Leaf Waxes and Hemicelluloses in Topsoils Reflect the δ2H and δ18O Isotopic Composition of Precipitation in Mongolia

Compound-specific hydrogen and oxygen isotope analyzes on leaf wax-derived n-alkanes (δ2Hn–alkane) and the hemicellulose-derived sugar arabinose (δ18Oara) are valuable, innovative tools for paleohydrological reconstructions. Previous calibration studies have revealed that δ2Hn–alkane and δ18Oara reflect the isotopic composition of precipitation, but – depending on the region – may be strongly modulated by evapotranspirative enrichment. Since no calibration studies exist for semi-arid and arid Mongolia so far, we have analyzed δ2Hn–alkane and δ18Oara in topsoils collected along a transect through Mongolia, and we compared these values with the isotopic composition of precipitation (δ2Hp–WM and δ18Op–WM, modeled data) and various climate parameters. δ2Hn–alkane and δ18Oara are more positive in the arid south-eastern part of our transect, which reflects the fact that also the precipitation is more enriched in 2H and 18O along this part of the transect. The apparent fractionation εapp, i.e., the isotopic difference between precipitation and the investigated compounds, shows no strong correlation with climate along the transect (ε2H n–C29/p = −129 ± 14‰, ε2H n–C31/p = −146 ± 14‰, and ε18O ara/p = +41 ± 2‰). Our results suggest that δ2Hn–alkane and δ18Oara in topsoils from Mongolia reflect the isotopic composition of precipitation and are not strongly modulated by climate. Correlation with the isotopic composition of precipitation has root-mean-square errors of 13.4‰ for δ2Hn–C29, 12.6 for δ2Hn–C31, and 1.2‰ for δ18Oara, so our findings corroborate the great potential of compound-specific δ2Hn–alkane and δ18Oara analyzes for paleohydrological research in Mongolia.


INTRODUCTION
Leaf wax-derived n-alkanes and hemicellulose-derived sugars are produced by higher terrestrial plants and stay wellpreserved in soils and sediments, because of their resistance against biochemical degradation (Eglinton and Hamilton, 1967). These compounds and their compound-specific stable hydrogen (δ 2 H n-alkane ) and oxygen (δ 18 O sugar ) isotopic composition get incorporated into soils through above-ground and root litter, abrasion, as well as grazing (dung), and they have a mean residence time of ∼40 years (leaf wax n-alkanes), while pentoses (including the hemicellulose-derived sugar arabinose) average over ∼20 years (Schmidt et al., 2011). Since δ 2 H n-alkane and δ 18 O sugar are not strongly affected by degradation effects (Zech et al., 2011(Zech et al., , 2012, they are increasingly used for paleohydrological reconstructions (Aichner et al., 2015;Hepp et al., 2015Hepp et al., , 2019Thomas et al., 2016;Rach et al., 2017;Schäfer et al., 2018;Bliedtner et al., 2020). Usually, they are interpreted to record the isotopic composition of precipitation (Sachse et al., 2006(Sachse et al., , 2012Tuthorn et al., 2015;Hou et al., 2018;Hepp et al., 2020), which in turn is long acknowledged as a valuable proxy for paleoclimate reconstructions and controlled by e.g., the temperature and amount effect, continentality, and altitude (Dansgaard, 1964). The isotopic signal of precipitation can be altered by isotopic fractionation at the soil-plantatmosphere interface, including evaporative enrichment of soil water (ε SW ), transpirative enrichment of leaf water (ε Et ) and biosynthetic fractionation (ε bio ) (Schimmelmann et al., 2006;Sachse et al., 2012;Liu et al., 2016;Cormier et al., 2018;Liu and An, 2019). ε SW and ε Et can lead to more positive δ 2 H n-alkane and δ 18 O sugar values relative to precipitation and/or the plants source water. ε SW is probably of minor importance under semi-arid and arid conditions because most plants (especially perennial plants) exploit deeper water sources that are not isotopically affected by evaporation (Feakins and Sessions, 2010;Kahmen et al., 2013a;Berke et al., 2015), while ε Et has a significant influence on δ 2 H n-alkane and δ 18 O sugar (Hou et al., 2008;Feakins and Sessions, 2010;Tuthorn et al., 2014Tuthorn et al., , 2015Berke et al., 2015;Cernusak et al., 2016;Liu et al., 2017).
So far, only very few calibration studies applied compoundspecific δ 18 O sugar analyzes Hepp et al., 2016Hepp et al., , 2020Strobel et al., 2020). For semi-arid and arid regions in South America and South Africa, Tuthorn et al. (2015) and Strobel et al. (2020) suggest enhanced evapotranspirative enrichment and higher ε 18O sugar/p values with increasing aridity.
By now, δ 2 H n-alkane and δ 18 O sugar topsoil calibration studies on modern reference material do not exist for semi-arid and arid Mongolia, so it is not clear whether δ 2 H n-alkane and δ 18 O sugar reflect the isotopic composition of precipitation and/or are strongly modulated by evapotranspirative enrichment and climate conditions. Therefore, the aim of this study is to evaluate the influence of different climatic factors on the isotopic composition of leaf wax-derived n-alkanes (δ 2 H n-alkane ) and the hemicellulose-derived sugar arabinose (δ 18 O ara ) in topsoils from semi-arid and arid Mongolia. More specifically, we addressed the following research questions: (1) Do δ 2 H n-alkane and δ 18 O ara reflect the isotopic composition of precipitation along the investigated transect?
Our study will provide the necessary basis for using δ 2 H n-alkane and δ 18 O ara for paleohydrological and -climatological reconstructions in semi-arid and arid Mongolia and similar regions.

Study Area and Sampling
The semi-arid and arid regions of Mongolia are influenced by three major atmospheric circulation systems ( Figure 1A). Summer climate is mainly dominated by the Westerlies and the East Asian summer monsoon (EASM) (Wang and Feng, 2013;Rao et al., 2015). Winter climate is dominated by the Siberian high blocking the Westerlies and thus moisture supply during winter (Yamanaka et al., 2007;Liu et al., 2009). This results in short and hot summers and long, cold, and dry winters and overall harsh conditions (Dashkhuu et al., 2015). Especially the vegetation period, during which biosynthesis can take place, is very short and corresponds with the summer months June, July, and August, when ∼75% of the annual precipitation occurs (Figures 1D,E, 2H; Lang et al., 2020).
The climate in Mongolia is characterized by increasing mean annual temperature (MAT) and decreasing mean annual precipitation (MAP) toward the south-east ( Figure 1C). This climate gradient is mirrored in regional vegetation biomes, welladapted to the semi-arid and arid conditions of Mongolia. Northern and central Mongolia are characterized by taiga, mountain-and forest steppe biomes, whereas steppe and desert steppe biomes are dominant in southern Mongolia (Hilbig, 1995;Klinge and Sauer, 2019). The precipitation shows a distinct 2 H p-OIPC and 18 O p-OIPC enrichment in southern and eastern arid Mongolia ( Figure 1A; Bowen, 2019), and the seasonal pattern is characterized by isotopically 2 H-and 18 O-depleted precipitation during winter and 2 H-and 18 O-enriched precipitation during summer (see Figures 1D For this study, topsoils (0-5 cm) were sampled in LDPE plastic bags in June/July 2017 (ID: 1-33) and June 2016 (ID: 34-42). To prevent molding, samples were stored open and dark during the 2-week field campaigns. Our sampling sites were dominated by different Poaceae and Cyperaceae species (grasses) as well as herbaceous and shrubby growth forms of Artemisia spp. and the woody shrub Caragana spp. Larix sibirica occurred at a few sites (for more details, see Struck et al., 2020). The investigated transect follows a west-east and north-south gradient with increasing aridity and 2 H p and 18 O p enrichment toward eastern and southern Mongolia (Figures 1A,B, 2 -all sampling sites and respective climate parameters are listed in the Supplementary Material).
While the n-alkanes had been extracted and quantified in a previous study  and were available and ready for compound-specific δ 2 H analyzes, we have selected 28 sampling sites for additional sugar analyzes. Hemicellulosederived sugars were hydrolytically extracted from 0.1 to 1.4 g topsoil material using 10 ml of 4M trifluoroacetic acid at 105 • C for 4 h according to Amelung et al. (1996). Thereafter, samples were vacuum-filtrated over glass fiber filters and the extracted sugars were cleaned over XAD-7 and Dowex 50WX8 columns to remove humic-like substances and cations (Zech and Glaser, 2009). The purified sugar samples were rotary-evaporated and derivatized with methylboronic acid (4 mg in 400 µl pyridine) at 60 • C for 1 h. α-Androstane and 3-O-Methyl-Glucose were used as internal standards (Zech and Glaser, 2009).

Stable Isotope Analyzes
The compound-specific hydrogen isotopic composition of the most abundant n-alkanes (n-C 29 , n-C 31 ) were measured on an Isoprime Vision isotope ratio mass spectrometer (IRMS) (Elementar, Langenselbold, Germany) coupled to an Agilent 7890B GC (Agilent, Santa Clara, United States) via a GC5 pyrolysis/combustion interface (Elementar, Langenselbold, Germany). The GC5 was operating in pyrolysis mode with a Cr (ChromeHD) reactor at 1050 • C. The GC was equipped with a split/splitless injector and an Agilent HP5GC fused silica column (30 m × 320 µm × 0.25 µm film thickness). Samples were injected in splitless mode and measured as triplicates. For normalization, n-alkane standards (n-C 27 , n-C 29 , and n-C 33 ) with known isotopic composition (Schimmelmann standard, Indiana) were measured as duplicates after every third sample triplicate. All measurements were drift-corrected relative to the standards in each sequence. The H + 3 -correction factor was checked regularly throughout the sequence and yielded stable values of 3.9 ± 0.02 (n = 4). The standard deviation of the sample triplicates was on average 1.2 and 1.0 for n-C 29 and for n-C 31 , respectively, and not worse than 3.6 and 4.7 . The standard deviation for all standards was better than 2 (n = 36). The hydrogen isotopic composition is given in delta notation (δ 2 H) vs. Vienna Standard Mean Ocean Water (VSMOW).
The compound-specific oxygen isotope measurements were performed on a Trace GC 2000 coupled to a Delta V Advantage IRMS using an 18 O-pyrolysis reactor (GC IsoLink) and a ConFlo IV interface (all devices from Thermo Fisher Scientific, Bremen, Germany). Samples were injected in splitless mode and measured in triplicates. For normalization, derivatized sugar standards with known isotopic composition were measured repeatedly at different concentrations within every sequence. Measured δ 18 O values were drift-and amount-corrected and corrected for the oxygen from the carbonyl group within the sugar molecules that became introduced during the hydrolysis according to Zech and Glaser (2009). The standard deviation of the sample triplicates was on average 0.4 , and not worse than 1.2 . The standard deviation of the arabinose standard was 1.8 (n = 24, average over four concentrations). Fucose and xylose concentrations were too low for robust isotope measurements; we therefore refrained from further evaluation of those data. The oxygen isotopic composition is given in delta notation (δ 18 O) vs. VSMOW.

DATA ANALYSIS
Apparent fractionation (ε app ) of hydrogen-and oxygen isotopes were calculated after Sauer et al. (2001) to test for climatic/environmental controls on δ 2 H n-alkane (Eq. 1) and δ 18 O ara (Eq. 2).  (Jarvis et al., 2008). Correlations of δ 2 H n-alkane and δ 18 O ara with δ 2 H p-WM and δ 18 O p-WM , respectively were tested using weighted linear regressions. Correlations of ε 2H n-alkane/p and ε 18O ara/p with climate were tested using unweighted linear regressions. Goodness of fit can be assessed using R 2 , and the accuracy using root-mean-square errors (RMSE). Differences between the arid part of the transect (ID: 34-42) compared to the rest were analyzed using a t-test. For data sets with unequal variance, the Welsh-corrected t-test was used. Statistical analyzes were done using the statistical software Origin 2019b.

Differences in Compound-Specific δ 2 H
Along the investigated transect δ 2 H n-C29 is on average ∼15 more positive than δ 2 H n-C31 . As described previously by Struck et al. (2020), n-C 29 is the most abundant homolog in the woody shrubs Caragana spp. and Artemisia spp., whereas n-C 31 is the most abundant homolog in grasses. Since shrubs and dicotyledonous plants in general are more sensitive to evapotranspirative enrichment than grasses (Sachse et al., 2012;Kahmen et al., 2013b;Hepp et al., 2020), the observed offset might indicate (i) plant-physiological differences affecting the evapotranspirative enrichment of different plants, and (ii) plantphysiological differences affecting ε bio .
Grasses grow via the intercalary meristem, where the leaf water is isotopically not as 2 H-enriched due to transpiration compared to the exposed part of grasses Ehleringer, 2000, 2002;Lehmann et al., 2017;Liu et al., 2017). The leaf wax n-alkanes, which are produced in the intercalary meristem, do therefore not incorporate the full leaf water enrichment signal (Barbour et al., 2004;Ripullone et al., 2008;Sachse et al., 2012;Kahmen et al., 2013b;Cernusak et al., 2016;Holloway-Phillips et al., 2016). This can be referred to as "dampening effect". δ 2 H n-alkane and δ 18 O ara Against the Isotopic Composition of Precipitation The δ 2 H n-alkane and δ 18 O ara values correlate significantly with the δ 2 H p-WM and δ 18 O p-WM values (Figure 3, R 2 = 0.30, p = 3.22e −4 for δ 2 H n-C29 ; 0.11 and 0.03 for δ 2 H n-C31 ; and 0.33 and 2.60e −4 for δ 18 O ara ). Significant correlations between compound-specific biomarker isotopes and the isotopic composition of precipitation have been observed previously for different regions (a.o.: Sachse et al., 2006;Feakins and Sessions, 2010;Hou et al., 2018;Li et al., 2019;Strobel et al., 2020). In comparison to transects covering larger climate gradients (a.o. Hou et al., 2018), our determination coefficients are small and explain only up to ∼30% of the variability. However, this is due to the fact that our transect covers only a small climate gradient. The RMSE is 13.4 for δ 2 H n-C29 , 12.6 for δ 2 H n-C31 and 1.2 for δ 18 O ara (Figure 3) and thus indicates that the biomarkers accurately record the isotopic composition of precipitation along our transect. There are several possible explanations for the observed scatter, including: (i) uncertainties related to the modeled OIPC-based isotopic composition of precipitation and the WorldClim 2.0 reanalysis FIGURE 4 | Apparent fractionation ε 2H n-C29/p , ε 2H n-C31/p and ε 18O ara/p from the topsoil transect through Mongolia plotted against T JJA , P JJA , Et 0 , AI (Fick and Hijmans, 2017;Trabucco and Zomer, 2019). Red trend lines illustrate linear regressions, gray shaded areas the 95% confidence interval. Bold R 2 /p-values indicate the level of significance (α = 0.05). Black dashed lines illustrate values for the biosynthetic fractionation, and black arrows indicate the effect of evapotranspirative enrichment (biosynthetic fractionation factors of −160 and +27 are assumed for the n-alkanes and arabinose, respectively).
Assuming constant ε bio values of −160 for leaf wax n-alkanes (Sessions et al., 1999;Hepp et al., 2020), evapotranspirative enrichment would be ∼31 for δ 2 H n-C29 and ∼15 for δ 2 H n-C31 (Figure 4). While n-C 29 is the most abundant homolog in shrubs, and n-C 31 is the most abundant homolog in grasses Struck et al., 2020) the observed offset in evapotranspirative enrichment most likely results from plant physiological differences described above, particularly the dampening effect. Assuming a constant ε bio factor of +27 for arabinose (Lehmann et al., 2017;Hepp et al., 2020) evapotranspirative enrichment would be ∼14 for δ 18 O ara . While we expect arabinose to be mainly synthesized by grasses (Mekonnen et al., 2019), we cannot quantify the contribution from other plants or roots (Schädel et al., 2010). Anyhow, for arabinose synthesized by grasses, we can expect a similar dampening effect as described above for leaf waxes, because leaf water in the leaf growth-and-differentiation zone is usually not as enriched as the exposed part of grasses (Lehmann et al., 2017;Liu et al., 2017;Hepp et al., 2020).

Comparison With Other Studies
While other calibration studies in relatively arid regions have also not found a strong climatic modulation of ε 2H n-alkane/p , our values (−129 ± 14 for ε 2H n-C29/p and −146 ± 14 for ε 2H n-C31/p ) are only comparable to those from Strobel et al. (2020), i.e., −133 ± 12 for n-C31 and n-C33, which are much more negative than those reported by Hou et al. (2008) and Feakins and Sessions (2010), i.e., −99 ± 8 and −92 ± 21 , respectively. Most likely, plant physiological and metabolic adaptations play an important role, as leaf waxes from C 4 plants are more enriched in 2 H than leaf waxes from C 3 plants, and dicotyledons produce more enriched leaf waxes than monocotyledons (Sachse et al., 2012;Kahmen et al., 2013b;Hepp et al., 2020). Li et al. (2019) reported very similar ε 2H n-C29/p values as for Mongolia from the semi-arid and arid regions in China (−127 ± 10 compared to −129 ± 14 ), yet ε 2H n-C31/p is much less negative than in Mongolia (−133 ± 13 compared to −146 ± 14 ). We suggest that this reflects the fact that C 4 grasses are more dominant in China than along our transect that is dominated by C 3 grasses (Pyankov et al., 2000). Our ε 2H n-C31/p values are also very much comparable to those reported for C 3 grass sites along a transect in Europe (Hepp et al., 2020), and our ε 2H n-C29/p values are in very good agreement with their ε 2H n-alkane/p values for sites dominated by deciduous trees, which ones more highlights the dampening effect of C 3 grasses compared to dicotyledons.
The ε 18O ara/p values for Mongolia (41 ± 2 ) are very similar to values reported by Strobel et al. (2020) for relatively arid regions in South Africa. There, the more humid regions have a significantly lower ε 18O ara/p (∼37 ), quite similar to the C 3 grass sites in Europe (Hepp et al., 2020). The deciduous tree sites in Europe, however, are again characterized by more enriched δ 18 O sugar values (ε 18O sugar/p = ∼43 ). All this indicates that δ 18 O is more sensitive to evapotranspirative enrichment than δ 2 H, so that climate can more strongly modulate δ 18 O sugar , and again that grasses show the signal dampening much more pronounced than dicotyledons.

CONCLUSION
This study investigated compound-specific δ 2 H n-alkane and δ 18 O ara values in topsoils collected along a transect through semiarid and arid Mongolia in order to evaluate to which degree they reflect variations in the isotopic signature of precipitation and/or they are affected by climate and evapotranspirative enrichment. We therefore tested for correlations of δ 2 H n-alkane and δ 18 O ara with δ 2 H p-WM and δ 18 O p-WM , respectively, as well as ε app with climate. We can conclude the following: • Leaf wax-derived n-alkanes and the hemicellulose-derived sugar arabinose are significantly more enriched in 2 H and 18 O in the more arid southern and eastern parts of the transect. This reflects the changes in the isotopic composition of precipitation along the transect, and the correlations with δ 2 H p-WM and δ 18 O p-WM have RMSE of 13.4 for δ 2 H n-alkane and 1.2 for δ 18 O ara . • The apparent fractionation remains mostly constant at −129 ± 14 , −146 ± 14 , and at + 41 ± 2 for ε 2H n-C29/p , ε 2H n-C31/p and ε 18O ara/p , respectively. There are no significant differences along the transect, nor strong correlations with climate.
Compound-specific δ 2 H n-alkane and δ 18 O ara analyzes on terrestrial biomarkers, preserved e.g., in lake sediments, have great potential for reconstructing past changes in the isotopic composition of precipitation and thus for paleoclimate andhydrological reconstructions in Mongolia.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in the article/Supplementary Material.