First Calibration and Application of Leaf Wax n-Alkane Biomarkers in Loess-Paleosol Sequences and Modern Plants and Soils in Armenia

Interpreting paleoenvironmental conditions by means of n-alkane biomarker analyses is challenging because results depend on different influencing factors. Thus, regional calibration of n-alkane patterns is needed because of different plant chemo-taxonomic behavior. We investigated for the first-time leaf wax-derived n-alkane biomarkers from modern plants, litter, top soils, and two recently discovered loess-paleosol sequences (LPSs) in Armenia (Lesser Caucasus). Our results on modern samples show a promising discrimination power based on n-alkane chain length nC33 (probably nC31)) for grasses and herbs versus nC29 for deciduous trees, despite the large interplant variability within vegetation groups. In contrast with other Loess records in Europe, where Late Pleistocene environments are ranging from tundra-like (glacial) to deciduous forest habitats (interglacial), our results from two Armenian LPSs suggest a transition from humid-steppe biome or forest-steppe vegetation dominating during interglacial periods, to semi-desert shrubs species more adapted to the enhanced aridity during glacial periods.


Introduction
Armenia has already reported evidence of climate change over the past years. Future climate projections predict that, because of global warming, the average temperature will further rise by 2-3 • C and rainfall will decrease by 10-15% within the next 50-100 years [1]. This continuous climatic aridification is likely to cause shifts in ecosystems affecting particularly species, which are least tolerant to drought [2] and cause significant negative impacts on agriculture and water resources. To improve our understanding of this critical environmental situation, and to predict the consequences of future climate change, it seems important to intensify investigations on paleoclimate changes and June, while the second half of the summer is dry. Low-land areas have mean annual precipitation (MPA) of between 200 and 250 mm, and higher altitudes receive precipitation of between 800 and 1000 mm [2]. Mountain steppes represent the dominant ecosystem of Armenia and occur throughout the country at altitudes between 1200 and 2000 m (sometimes as high as 2500 m). In comparison with other countries of the Lesser Caucasus region, Armenia has fewer forests. Less than 10% of its territory is covered by forests, mostly concentrated in the North-East (62%) and in the South [30], while only 2% corresponds to the central area and the Lake Sevan Basin [2]. Armenian forests are predominantly broadleaved (98.6%), dominated by oak, beech, and hornbeam trees. Dry scrub forests are to be found at altitudes of between 900 and 1000 m in northern parts of the country but occur at much higher altitudes in the southern parts (1800-2000 m). These forests harbor around 80 species of xeric trees and shrubs, all of which are drought tolerant and light-loving [31,32].
The studied loess-paleosol sequences (LPSs) are situated in the northeastern part of Armenia, at the northern flanks of the Lesser Caucasus. All sections are located close to the town of Sevkar north of Ijevan the capital of Tavush province (41 • N, 45 • 10 E; see Figure 1) at an altitude of between 680 and 960 m a. s. l. These loess deposits form elongated ridges up to 14 m thick, 150 m wide, and 1-2 km long [7].
Geosciences 2019, 9, x FOR PEER REVIEW 3 of 17 (MPA) of between 200 and 250 mm, and higher altitudes receive precipitation of between 800 and 1000 mm [2]. Mountain steppes represent the dominant ecosystem of Armenia and occur throughout the country at altitudes between 1200 and 2000 m (sometimes as high as 2500 m). In comparison with other countries of the Lesser Caucasus region, Armenia has fewer forests. Less than 10% of its territory is covered by forests, mostly concentrated in the North-East (62%) and in the South [30], while only 2% corresponds to the central area and the Lake Sevan Basin [2]. Armenian forests are predominantly broadleaved (98.6%), dominated by oak, beech, and hornbeam trees. Dry scrub forests are to be found at altitudes of between 900 and 1000 m in northern parts of the country but occur at much higher altitudes in the southern parts (1800-2000 m). These forests harbor around 80 species of xeric trees and shrubs, all of which are drought tolerant and light-loving [31,32]. The studied loess-paleosol sequences (LPSs) are situated in the northeastern part of Armenia, at the northern flanks of the Lesser Caucasus. All sections are located close to the town of Sevkar north of Ijevan the capital of Tavush province (41°N, 45°10'E; see Figure 1) at an altitude of between 680 and 960 m a. s. l. These loess deposits form elongated ridges up to 14 m thick, 150 m wide, and 1-2 km long [7]. The present climate of the loess area is characterized by annual precipitation of between 450 and 550 mm, and a mean annual temperature of about 11 °C. Nowadays, the area is intensively used for agriculture. However, the potential natural vegetation according to recent climate points to forest associations (mesophytic deciduous broad-leaved and mixed coniferous-broadleaved forests) [33]. Accordingly, mesophilic and diversely mixed forests set in as soon as the relief becomes too steep for agricultural activity. A few kilometers to the north, the flanks of the Lesser Caucasus change over into the Transcaucasian depression, a lowland basin that is drained by the Kura River. This transition is related to a natural vegetation change leading over to steppic and dwarf semi-shrub desert plant communities [33]. A spatial shift of this vegetation line, especially during moister or dryer climatic conditions of the late Quaternary period should be considered.
In order to identify late Quaternary vegetation changes in our study area, we investigated two LPSs based on the analysis of leaf-wax derived n-alkane patterns. For getting a key to interpret these n-alkane patterns, we established a modern reference data set of n-alkane patterns. Therefore, we collected 40 samples of fresh plant material as well as six litter samples and six topsoil samples (0-5 cm). Top soils were collected as mixed samples from below the dominant vegetation (2 in grasslands The present climate of the loess area is characterized by annual precipitation of between 450 and 550 mm, and a mean annual temperature of about 11 • C. Nowadays, the area is intensively used for agriculture. However, the potential natural vegetation according to recent climate points to forest associations (mesophytic deciduous broad-leaved and mixed coniferous-broadleaved forests) [33]. Accordingly, mesophilic and diversely mixed forests set in as soon as the relief becomes too steep for agricultural activity. A few kilometers to the north, the flanks of the Lesser Caucasus change over into the Transcaucasian depression, a lowland basin that is drained by the Kura River. This transition is related to a natural vegetation change leading over to steppic and dwarf semi-shrub desert plant communities [33]. A spatial shift of this vegetation line, especially during moister or dryer climatic conditions of the late Quaternary period should be considered.
In order to identify late Quaternary vegetation changes in our study area, we investigated two LPSs based on the analysis of leaf-wax derived n-alkane patterns. For getting a key to interpret these n-alkane patterns, we established a modern reference data set of n-alkane patterns. Therefore, we collected 40 samples of fresh plant material as well as six litter samples and six topsoil samples (0-5 cm). Top soils were collected as mixed samples from below the dominant vegetation (2 in grasslands and 4 in sites with deciduous trees/ shrubs). Most sampling sites were situated nearby the studied LPSs in northeastern Armenia, and in the region of Maghavuz in the Republic of Artsakh (Nagorno-Karabakh) where new loess sections were recently discovered. Our approach was to sample plants that are very widespread nowadays, with the assumption that they may have been widespread during similar climate phases in the past as well. As coniferous trees are well known to produce very low n-alkane concentrations, which do not contribute significantly to paleo-records [28,34,35], we refrained from sampling and further considering coniferous trees in our approach. For the studies on the LPSs, two loess sections with a thickness of 28 m (Border line (BL) section) and 21 m (Achajur (AJ) section) were selected to carry out n-alkane analysis ( Figure 2). In total, we took 120 samples from the first section (BL) to include all relevant layers and soil horizons that represent at least three glacial-interglacial cycles [7]. Moreover, we took 40 samples from the well-differentiated upper part of the second profile (AJ) that particularly reflects the last glacial period. and 4 in sites with deciduous trees/ shrubs). Most sampling sites were situated nearby the studied LPSs in northeastern Armenia, and in the region of Maghavuz in the Republic of Artsakh (Nagorno-Karabakh) where new loess sections were recently discovered. Our approach was to sample plants that are very widespread nowadays, with the assumption that they may have been widespread during similar climate phases in the past as well. As coniferous trees are well known to produce very low n-alkane concentrations, which do not contribute significantly to paleo-records [28,34,35], we refrained from sampling and further considering coniferous trees in our approach. For the studies on the LPSs, two loess sections with a thickness of 28 m (Border line (BL) section) and 21 m (Achajur (AJ) section) were selected to carry out n-alkane analysis ( Figure 2). In total, we took 120 samples from the first section (BL) to include all relevant layers and soil horizons that represent at least three glacialinterglacial cycles [7]. Moreover, we took 40 samples from the well-differentiated upper part of the second profile (AJ) that particularly reflects the last glacial period.

Stratigraphic Patterns of the Loess-Paleosol Sequences
A detailed description of the BL section is already published in [7]. To introduce the new Achajur (AJ) section, a brief description of the stratigraphic succession is given below (see Figure 5). The stratigraphic order of the sedimentation units and the pedo-complexes in the two profile sections were quite similar. They were subdivided into different sequences, which could incorporate loess material, loess-derived soils, soil complexes, colluvial layers, and tephra. From the bottom to the top we differentiated: • Sequence 5 (17.7-28.5 m in BL profile) Sequence 5 starts with a thick succession (about 10 m) of strongly weathered dark-brown loams. The occurrence of strong enrichment of secondary carbonate indicates in situ soil formation in the depth of 23.5-26.7 m and 19.8-21.2 m. Each soil is linked to clay contents between 40% and 60% and strong reddish-dark brown to blackish-brown colors. The top of the sequence (until 17.7 m) is characterized by the formation of a blackish soil with high clay contents (>50%), high organic carbon contents (~0.8%), and intense illuviations into drying cracks or root channels.
• Sequence 4 (16-17.7 m in BL profile) Sequence 4 shows neither strong weathering nor intense loess accumulation. Instead, alternating phases of sediment redistribution and subsequent soil formation occurred, leading to blackish horizons rich in organic carbon (pedocomplex (P-3), after Wolf et al. [7]. In the AJ profile, a definite

Stratigraphic Patterns of the Loess-Paleosol Sequences
A detailed description of the BL section is already published in [7]. To introduce the new Achajur (AJ) section, a brief description of the stratigraphic succession is given below. The stratigraphic order of the sedimentation units and the pedo-complexes in the two profile sections were quite similar. They were subdivided into different sequences, which could incorporate loess material, loess-derived soils, soil complexes, colluvial layers, and tephra. From the bottom to the top we differentiated: • Sequence 5 (17.7-28.5 m in BL profile) Sequence 5 starts with a thick succession (about 10 m) of strongly weathered dark-brown loams. The occurrence of strong enrichment of secondary carbonate indicates in situ soil formation in the depth of 23.5-26.7 m and 19.8-21.2 m. Each soil is linked to clay contents between 40% and 60% and strong reddish-dark brown to blackish-brown colors. The top of the sequence (until 17.7 m) is characterized by the formation of a blackish soil with high clay contents (>50%), high organic carbon contents (~0.8%), and intense illuviations into drying cracks or root channels.
• Sequence 4 (16-17.7 m in BL profile) Sequence 4 shows neither strong weathering nor intense loess accumulation. Instead, alternating phases of sediment redistribution and subsequent soil formation occurred, leading to blackish horizons rich in organic carbon (pedocomplex (P-3), after Wolf et al. [7]. In the AJ profile, a definite allocation of Sequence 4 seems too uncertain at the moment.
• Sequence 3 (11.5-16 m in BL profile; 14.7-20 m in AJ profile) Sequence 3 represents a succession of loess deposits for both profiles, ending up with the formation of another pedocomplex (P-2). Below the P-2 pedocomplex, another weakly developed brownish soil is interposed between the loess layers in both sections. The pedocomplex (P-2) extends from 11.8 to 14.2 m in the BL section and from 15.2 to 18.5 m in the AJ section. The lower part of P-2 starts with a strong blackish-brown soil and pedogenic carbonate enrichment that is more intensely developed in the BL profile. In the upper part of P-2, the BL profile shows a black soil with low clay contents but the highest content of organic carbon in the whole sequence. In contrast, the soil in the upper P-2 in the AJ profile is characterized by strongly weathered reddish-brown material; however, the contribution of slope processes is indicated by numerous clasts and rock fragments. Sequence 3 ends up with a greyish colluvial layer including dark relocated soil material (clay pebbles) that covers the pedocomplex P-2.
• Sequence 2 (5.3-11.5 m in BL profile; 6-14.7 m in AJ profile) Sequence 2 is characterized by the deposition of thick layers of loess material with low clay contents and bright colors pointing to just limited weathering processes. A pale blueish tephra layer (9.7-10.1 m in BL profile; 12.5-12.9 m in AJ profile) interrupts these loess layers. While in the BL profile the tephra is covered by homogeneous ochre-colored primary loess, the AJ profile clearly points to synor post-sedimentary relocation processes that are evidenced by small rock fragments and numerous thin layers of reworked tephra material. In the upper part of sequence 2, a pedocomplex P-1 (Sevkar soil complex) has been formed during marine isotope stage (MIS) 5 as indicated by first OSL-dating [7]. P-1 extends from 5.3 to 7.9 m in BL profile and 6.2 to 9.5 m in AJ profile. It consists of a succession of three palaeosoils and shows a transition from brown clayey soils in the lower part to more blackish and crumby soils in the upper part, which seems quite comparable with the pedocomplex P-2 in the BL profile.
• Sequence 1 (0-5.3 m in BL profile; 0-6m in AJ profile) Sequence 1 is characterized by a succession of loess deposits with interposed palaeosoils and has been formed during the last glacial (MIS 4 to MIS 2). This sequence shows a very different appearance in both profiles. In general, the loess material seems to be more weathered comparing to the layers below. But in the BL profile, this sequence is divided by two relatively strongly weathered brown soils, while the AJ profile is characterized by up to five weakly-developed and less-intense colored palaeosoils. So far, and without the support by micromorphological analyses, we are not able to identify these soils without a doubt. But it is already evident that, in contrast to the BL profile that experienced most probably two pronounced stability periods during the last glacial, the AJ profile was characterized by stronger sedimentation and relocation dynamics and influenced by less stable conditions. At the top of the sequence, a Holocene soil (P-0) has been developed (0.5-1.5 m in BL profile and 1.3-1.9 m in AJ profile) that is covered by Holocene colluvial deposits in both sections.

Leaf Wax Analyses and Calculation of n-Alkane Indices
Leaf wax analyses were carried out in the laboratories of the Soil-Biogeochemistry Group of the Martin Luther University in Halle-Wittenberg. Total lipids from modern plants, top soils, and loess sections were extracted with organic solvents DCM:MeOH (9:1), using Soxhlet apparatuses for 24 h. After concentration by rotary evaporation, each total lipid extract was passed over a pipette column filled with aminopropyl silica gel. The aliphatic fraction, containing the n-alkanes, was eluted with 3 mL n-hexane, whereas the polar fraction (e.g., alcohols) and the acid fraction were obtained by DCM:MeOH (1:1) and 5% acid in diethyl ether, respectively. Quantification of the n-alkanes was performed on a gas chromatograph coupled to a flame ionization detector (GC-FID). External standards (nC 8 -nC 40 ) were run with each sequence for identification and quantification.
Total n-alkane concentration (TAC) in plant and soil samples was calculated as the sum of individual n-alkane concentrations from nC 25 to nC 35 (odd as well as even ones) and was given in µg/g dry weight (dw).
Odd-over-even predominance (OEP) was used as a proxy for degradation and was determined after Hoefs et al. [36]. Low values (<5) indicated an enhanced state of degradation [8,16].
The average chain length (ACL) of n-alkanes, which is the weighted average of the different carbon chain lengths, was calculated by modifying the equation of Poynter et al. [37].

Total n-Alkane Concentrations (TAC), Indices (OEP, ACL), and Patterns of Modern Plants
The investigated samples showed total n-alkane concentrations between 29 and 2655 µg/g plant dry weight (dw) (Figure 3a). These values were in the lower to middle range of the interval that has been reported for n-alkane concentrations of herbaceous and woody plant species worldwide (0.4-7115 µg/g dw [27,28,34,[38][39][40]). Grasses and herbs produced slightly higher amounts of n-alkanes with a mean value of 666 µg/g dw compared to deciduous trees and shrubs, with a mean TAC value of 491 µg/g dw (see Supplementary Materials). For deciduous trees, the highest n-alkane concentration was produced by Hawthorn trees (Crataegus spp.; 2655 µg/g dw) followed by oak trees (Quercus spp.; 1359 µg/g dw) and beech trees (Fagus spp.; 651 µg/g dw). The latter two species were the most common ones in Armenian forest communities. The long-chain n-alkanes of all investigated modern plant samples revealed a strong odd-over-even predominance (OEP). This is typical for epicuticular leaf waxes of terrestrial higher plants [22]. For deciduous trees and shrubs, about 50% of the samples showed an OEP ranging from 11 to 20, with a minimum value of 8 and a maximum of 30. Grasses and herbs showed quite similar middle values ranging between 9 and 17, with a minimum of 8 and a maximum of 38 ( Figure 3b). The OEP is generally considered as a proxy for the preservation status of the leaf-wax derived n-alkanes [41][42][43][44], with high values being characteristic of not-or just little-degraded n-alkanes [34,45]. Thus, OEP values from Armenian plant samples were typical for fresh plant material. The average chain lengths (ACL) of the n-alkanes indicated a little difference between the two vegetation groups: Grasses and herbs yielded a slightly higher median value compared to trees and shrubs ( Figure 3c). Moreover, nC 29 and nC 31 tended to be the most abundant n-alkanes in all investigated plant samples (Figure 4a). Our results showed that grasses and herbs generated an equal predominance of both nC 29 and nC 31. However, deciduous trees produced a slightly higher proportion of nC 29. In contrast, nC 33 did not represent a dominant pattern but, overall, was produced in higher amounts by grasses and herbs. In different regions, a predominance of the chained homologs nC 31 and nC 33 has been found to be related with herbaceous species [46,47], whereas a higher abundance of the homologs nC 27 and nC 29 would be associated with deciduous trees and shrubs [46][47][48].
all investigated plant samples (Figure 4a). Our results showed that grasses and herbs generated an equal predominance of both nC29 and nC31. However, deciduous trees produced a slightly higher proportion of nC29. In contrast, nC33 did not represent a dominant pattern but, overall, was produced in higher amounts by grasses and herbs. In different regions, a predominance of the chained homologs nC31 and nC33 has been found to be related with herbaceous species [46,47], whereas a higher abundance of the homologs nC27 and nC29 would be associated with deciduous trees and shrubs [46][47][48]. and predominated over the other n-alkanes (Figure 4b). In addition, nC33, which is indicative of grasses and herbs as discussed above, tended to be higher in grassland top soils, too (Figure 4b).
Thus, there is a tendency to distinguish top soils of grasslands from top soils of deciduous forest sites by a slight predominance of nC31 compared to nC29, higher amounts of nC33, and thus slightly higher ACLs values in general (Figure 3c). Such an observation offers a promising base to distinguish between sites that were dominated by grasses and herbs or trees and shrubs (Figure 3d, Figure 4). However, the differences are not high, and this relation should be confirmed by enlarging the data set.  Similar patterns are also reported by a recent study conducted in Eastern Georgia [49], a region that is adjacent to our study area. However, such a clear chemotaxonomic character in n-alkanes production was not detectable in Armenian modern plant samples. Indeed, likewise in Armenia differences based on C 33 vs. C 29 appeared to be a promising tool for differentiating between grasses/ herbs and deciduous trees/shrubs, but this relation is still not confirmed by each sample. This might point to some interplant variability in n-alkane production within the same vegetation groups (Figure 3d). Furthermore, we assume that interplant variability and environmental stress may be responsible for the different n-alkane patterns between Armenia and Georgia. Nowadays, the climate in NE-Armenia is relatively drier, which represents a stress factor to vegetation and may be expressed in deviating n-alkane production. The same observation, albeit with less intensity, has been made in Georgia, where some local hydroclimatic effects and changes in humidity and temperature altered the n-alkane patterns of some plants [49]. Indeed, it has been repeatedly discussed in the literature [28,34,50,51] that local environmental conditions and related physiological stress could influence the patterns of n-alkane production between members of vegetation groups and within the same species, too. Leaf waxes are the plant's first barrier to the atmosphere. Thus, it is reasonable to hypothesize that their composition, including n-alkanes, could be controlled to some degree by environmental adaptation [28,52]. Some species exhibit a high degree of genetic controls over their n-alkane patterns, such as the almost exclusive production of nC 27 by Fagus sylvatica leaves [28,[53][54][55][56], a pattern also confirmed by our analyses to some extent (see Supplementary Materials). In contrast, leaf wax patterns of other plant species may be strongly altered by local environmental parameters, for example temperature and precipitation [37,38,42,56], radiation, nutrient and water availability, salinity, or pollution [52,57].
In a recent study, covering a large set of modern vegetation n-alkane data, Bush and McInerney [28] highlighted that the discriminant power based on the chain lengths nC 27 and nC 29 vs. nC 31 and nC 33 is never better than 75% for any geographical position. For the temperate zone, which also includes Armenia, they even showed that the discriminant power remains only correct for about 49% of the studied woody angiosperm species [28]. Actually, the temperate zone is strongly divided into smaller climate zones, with more variety in temperature and humidity over the year and stronger seasonality changes depending on continentality and highland-lowland effects. In that way, plants may record higher annual and/or monthly variability in produced n-alkane patterns, which possibly is demonstrated by the large standard deviations of the n-alkanes from modern plant material in Armenia. In view of this high variability within fresh plant material, the analysis of topsoil samples may have the advantage of integrating over the whole ecosystem and implicitly considering such variability [27].

Total n-Alkane Concentrations (TAC), Indices (OEP, ACL), and Patterns of Litter and Topsoil Samples
The total n-alkane concentrations for litter and topsoil samples were 79% and 98% lower, respectively, than those of fresh plant material (Figure 3a). OEP values declined notably from modern vegetation to litter and top soils (Figure 3b), and the ACLs tended to shift towards shorter values (Figure 3c). The considerable decrease in leaf wax n-alkane concentrations and OEP values in top soils as compared to fresh plant material were to be expected and indicated effects of degradation and microbial reworking of organic matter (OM) [8,27,58]. The n-alkane distribution of forest top soil samples was quite similar compared to leave patterns of deciduous trees and shrubs with both nC 29 and nC 31 dominating the sites (Figure 4b). However, in grassland top soils, nC 31 slightly increased and predominated over the other n-alkanes (Figure 4b). In addition, nC 33 , which is indicative of grasses and herbs as discussed above, tended to be higher in grassland top soils, too (Figure 4b). Thus, there is a tendency to distinguish top soils of grasslands from top soils of deciduous forest sites by a slight predominance of nC 31 compared to nC 29 , higher amounts of nC 33 , and thus slightly higher ACLs values in general (Figure 3c). Such an observation offers a promising base to distinguish between sites that were dominated by grasses and herbs or trees and shrubs (Figure 3d, Figure 4). However, the differences are not high, and this relation should be confirmed by enlarging the data set.
The last glacial loess and the soil complex P-1 showed slightly higher n-alkane concentrations. An increase in TAC values occurred with the changeover of soil properties within P-1, leading from a brown soil, towards a black-brown soil, and finally to a pure black soil. This might point to a higher bio-productivity or preservation under steppic conditions (Figures 5b and 6b). OEPs along the profiles showed a varying degree of preservation, with lower values pointing to enhanced degradation in some paleo-units. However, there was no general correlation between OEP and TAC. This could be explained by TAC being controlled by both n-alkane production and degradation, whereas OEP is primarily controlled by degradation only.
A clear synchronic fluctuation of the average chain length (ACL) and the n-alkane ratios nC 33 /(nC 29 +nC 33 ) and (nC 31 +nC 33 )/(nC 27 +nC 29 +nC 31 +nC 33 ) took place along the studied profiles (Figures 5d and 6d). High values that refer to the predominance of the homologs nC 31 and nC 33 were attributable to P-0, P-1, P-2, P-3, a strong blackish soil in the depth of 20 m (BL), and a strongly weathered Kastanozem-like soil at the depth of 23 m (BL). In contrast, a lower ACL and decreased n-alkane ratios pointed to the predominance of nC 29 , which can be mainly associated with glacial periods (e.g., at the depth of 2-5 m, 8-11 m, and~16 m).

Paleoenvironmental Reconstruction
As a general tendency in the production of n-alkane patterns, the chain lengths nC 31 and/or nC 33 are assumed to be predominantly produced by grasses and herbs, while nC 27 and/or nC 29 are mainly formed by trees and shrubs. This is suggested by several studies that have been especially undertaken in Central Europe [16,27,59]. Moreover, this discrimination is confirmed by modern n-alkane data sets from eastern Georgia as well as promising observations in the study in hand. If we now hypothesize that the pattern in Armenian LPSs follows the same rule, the changes in ACL and the n-alkane ratios would suggest that grasses and herbs dominated the ecosystem during the formation of the pedo-complexes (inter-glacial periods), whereas trees and shrubs mainly prevailed during glacial periods (Figures 5 and 6). However, below pedocomplex P-1, concentrations, as well as OEP values, were very low in certain paleo-units. Therefore, it must be considered that just weak changes (e.g., by OM degradation) may result in big changes in the displayed curves.
Although n-alkanes are chemically quite inert, various degrading bacteria can utilize them as carbon and energy sources. Thus, the detected signal is probably prone to degradation. Such observations were also discussed by other studies, which highlighted considerable effects of the ongoing degradation of OM in sediments and soils [8,16,17]. In order to prevent misinterpretations and ensure a reliable reconstruction of paleo-vegetation, it seems crucial to minimize possible effects that arise from degradation processes, which could lead to preferential losses of the most dominant alkane homologs. To this objective, Zech et al. [12,16,61] described a correction procedure ("end member model") in more detail. Based on a modern n-alkane data set of plant, litter, and topsoil samples from central and southeastern Europe, the model compares the normalized n-alkane ratio (nC 31 +nC 33 ) / (nC 27 +nC 29 +nC 31 +nC 33 ) with the OEP to differentiate between vegetation groups and integrate the degradation impact at the same time. In brief, "degradation lines" for n-alkanes derived from grasslands and forests are calculated by plotting n-alkane ratios against OEP values. Fossil sediments or soil samples formed under grasslands and deciduous forests should plot along the degradation lines of samples from modern data. Furthermore, the contribution of grasses/herbs versus deciduous trees/shrubs to the n-alkane signal can be estimated. The percentage of grass vegetation can be calculated by the following formula: % grass = n − alkane ratio (sample) − equation (degradation line trees) equation (degradation lines grass) − equation (degradation lines trees) * 100 Grass percentages that exceed 100% illustrate the uncertainties and limitations of the end-member modeling approach because percentages are just semi-quantitative estimations.
Bliedtner et al. [49] integrated n-alkane results of modern plant and topsoil samples collected from eastern Georgia into the dataset of Central Europe [59] and recalculated the degradation lines. They further suggest that these adjusted equations could be used to calculate the relative vegetation contributions of grasses vs. deciduous trees in paleo-records from the southern Caucasus and close regions.
After correction, the percentage of grass/herb-derived n-alkanes for both the central-European and the adjusted Caucasus models were calculated. (Figures 5e and 6e). Even after correcting the results by taking into consideration OM degradation effects, no major changes were found. Both models showed that it is very likely that herbaceous species were not the predominant source producing n-alkanes during the glacial/stadial periods and that deciduous trees and shrubs did not contribute significantly to n-alkanes patterns extracted from paleosols. Such findings do not match with the traditional paradigm of the treeless cold steppic environment during glacial conditions and developed deciduous forests in interglacial periods.

"Vegetation Source" versus Environmental Control of Detected n-Alkane Signals
In the same line with recent vegetation patterns, fossil n-alkane patterns may also have been controlled by stressful local environmental conditions (e.g., aridity, temperature). Such effects are widely observed in literature and likewise confirmed by the analysis of modern plant material in Armenia. Rinna et al. [61] demonstrated, in a study conducted on marine sediments from the Atlantic and Pacific ocean, that a possible relationship between fossil n-alkane chain lengths and latitudinal/altitudinal factors exists. They furthermore suggested that chain lengths of nC 31 , nC 33 , and nC 35 correlate with warmer and drier, or possibly more irradiated environments. In that way, patterns from pedo-complexes in Armenian LPSs pointing to the dominance of nC 31 and nC 33 chain lengths might be a response to a stressful environment (e.g., pronounced irradiation or very dry warm summers), boosting the production of the longer n-alkane chain lengths nC 31 and nC 33. Likewise, it may be expected that, during glacial conditions, a shortening of the n-alkane chain length produced by grasses and herbs may occur due to a decrease in temperature and the enhanced aridity [62]. Thus, the detected fossil n-alkane signal could reflect environmental conditions rather than the vegetation source itself.
However, the question remains to which extent stressful environments may globally control the production of n-alkane pattern and change plants chemotaxonomic characters based on the predominant formation of nC 27 and nC 29 by trees and shrubs and of nC 31 and nC 33 by grasses and herbs (see, e.g., [63]).

Deciduous Trees Refugia during Glacial Periods
According to the observed n-alkane patterns in the LPSs of NE-Armenia, one could argue that glacial periods in our study area have been linked to tree vegetation, while interglacial periods were characterized by treeless vegetation. Zech et al. [12] experienced similar n-alkane patterns in LPSs from the southeastern Carpathian Basin. They attributed such findings to ecologically-effective moisture, suggesting that low soil humidity, as a result of high evapotranspiration during interglacial periods, limited the survival of trees. In contrast, reduced evapotranspiration occurred under cooler climatic conditions favoring tree growth during glacial periods. Thus, it is assumed that parts of southeastern Europe may have served as refugia for deciduous trees during the glacial periods [63,64]. One pivotal question is whether such hydrological relations would be conceivable with respect to our study area in NE-Armenia as well. In general, the interpretation of higher moisture during glacials and drier conditions during interglacials strongly contradicts our interpretation based on stratigraphic and pedogenic results. Based on geomorphological principles, thick unaltered loess layers rather indicate high aridity, while strongly weathered paleosoils that are rich in clay and humic substances evidence more humid conditions. Further environmental evidence may arise from a comparison with other studies from a regional context. However, such studies are rare in the Caucasus region. Some pollen studies that have been conducted on Early to Middle Quaternary deposits (e.g., Shamb section-paleo-lake sediments in southern Armenia: [3]; North Caucasus and Kazakhstan: [65]) have suggested that the Pleistocene climate was characterized by an alternation of wet conditions during glacial periods and dry conditions during interglacial periods. Other, more modern pollen studies based on precise chronologies (Shamb section in southern Armenia: [5]; Dmanisi (Georgia): [66]; Lake Urmia (NW-Iran): [67]; Lake Van (Eastern Turkey): [68,69]) contradict this interpretation and suggest that glacial periods have been dry and cold and predominantly linked to a vegetation cover composed of grasses and herbs. In view of the high heterogeneity of hydrological conditions during glacial-interglacial cycles that must be expected for the mountainous Caucasus region, a comparison with these faraway paleoenvironmental archives will probably not lead to reliable statements. For this purpose, the analysis of a considerably closer Quaternary archive, such as the Lake Sevan in central-eastern Armenia would be needed. This also means that, for the moment, we have hardly any comparative information and all indications regarding paleoenvironmental conditions originate from the LPSs themselves.
One important hint on vegetation cover appears in the lowermost paleosoil of pedocomplex (P-1) that corresponds to the last interglacial period. Here, up to 5 cm thick root channels have been refilled with clay-humic material indicating the prevalence of tree vegetation. Moreover, the largely unaltered loess units contain no expressions of higher moisture during glacial periods. Low clay contents, low organic contents and bright colors [7] point to mostly absent weathering dynamics that, in turn, indicates very dry environmental conditions with strongly reduced soil humidity. Since such results strongly challenge the interpretation based on n-alkane patterns, more future activities are required to derive independent proxy data from the LPSs in order to achieve a valid reconstruction of paleoenvironmental conditions.

Grasses and Herbs as Indicators of Humid Conditions Rather Than of Drier Ones
A closer look to the pedocomlex (P-1) reveals that total n-alkane concentrations increased in correlation with a changeover of soil properties. Starting from a coarsely aggregated brown soil, leading to a black-brown soil, and finally to a crumbly pure black soil (steppe soil), an increasing TAC points to higher production and/or better preservation of alkanes under steppic conditions. The fact that appearance and characteristics of the brown paleosoils at the base of P-1 give evidence of intense weathering fits well with the indication of stronger alkane degradation in these paleosoils. However, the ACL did not change noticeably within the pedocomplex. Since the ACL and further alkane ratios indicate grass-dominated vegetation cover across all soils of the pedocomplex, the following conclusions may be drawn. As soon as environmental conditions became slightly wetter, incipient pedogenesis has been accompanied by the establishment of steppe vegetation that is indicated by the increase of grass-derived n-alkanes. Probably there was hardly any forest composed of deciduous trees, instead, high grasses dominated. For the last interglacial period, certain indications of tree vegetation were detected such as the filled root channels, but the observed n-alkane signal still refers to predominant grass vegetation. A forest steppe environment may explain such findings because the biomass of grasses and herbs might be remarkably higher thanks to a longer growing season and higher growth rates compared with biomass produced by trees [63]. In the same direction, likewise, pollen analyses that were conducted in sediments of the Shamb paleo-lake in southern Armenia [5], demonstrate that steppe elements were always abundant at Shamb, even during interglacial periods and nowadays.
Later, when it became drier during glacial/stadial periods, a transition from humid grass biome to semi-desert shrub communities may have taken place, leading to the production of alkanes with the dominant chain length nC 29 (probably nC 27 , too). An ecological state transition from grasslands to dominated shrublands ecosystem are very common in arid regions nowadays (in Armenia too) and are usually attributed to various factors relating to pronounced aridity and limited precipitation [70][71][72]. With increased aridity there is a competition between grasses and shrubs. Certain shrubs have the ability to perform photosynthesis for a longer period of the growing season thanks to deep and laterally-extensive root systems that access deeper groundwater [73]. This limits the establishment of grass vegetation and promotes the growth of semi-desert shrub species. However, lower temperatures during glacial periods may have exerted considerable stress on shrubs growth, but probably positive temperature feedbacks may have occurred. Some modern research in the Chihuahuan Desert grasslands in the USA focusses on grasses-shrubs transition nowadays and demonstrates an increase of bare soils in shrub biomes. Resulting larger open spaces absorb greater amounts of solar radiation, which is transformed into thermal energy and stored in the ground, and later warms up the near-surface air layer close to the shrublands [74,75]. The authors show that increased nocturnal air temperatures may reduce juvenile mortality and favor continuing shrub growth and survival during cold periods [74,76].
To summarize, patterns of n-alkane production in NE-Armenian LPSs might indicate a transition from humid high steppe biome or forest steppe occurring during periods of soil formation (interglacial/interstadial) to semi-desert shrub species more adapted to enhanced aridity during glacial periods. This also implies that n-alkane-based grass signals may be seen as an indication of more humid environmental conditions, rather than of dry conditions as widely supposed.

Conclusions
Within this study, n-alkane biomarker analyses were applied to modern plants and top soils, as well as two recently discovered LPSs in NE-Armenia for the first time. Investigations on modern samples demonstrate that, despite the large interplant variability within vegetation groups that is mainly attributed to changes in local environmental conditions (e.g., changes in temperature and aridity), a promising discrimination power based on nC 33 (probably nC 31 as observed in grassland top soils) for grasses and herbs versus nC 29 for deciduous trees is detected.
The results from Armenian LPSs suggest three hypotheses: (1) In the same line with recent observations, stressful environments could globally control the production of n-alkane patterns and alter the plant's chemotaxonomic character. On the one hand, pronounced irradiation or very dry warm summers might boost the production of the n-alkanes chain length nC 31 and nC 33 , even by deciduous trees during interglacial periods. On the other hand, a shortening of the n-alkane chains produced by grasses and herbs might occur due to the enhanced aridity and temperature decrease during glacial/stadial periods. (2) The study area may have been an important glacial refuge for deciduous trees leading to tree survival even during glacial periods thanks to higher ecologically effective moisture (i.e., including effects of reduced evapotranspiration). However, recent pollen analysis from a wider Caucasian region together with certain soil and sediment features in Armenian LPS challenge such an interpretation. (3) A transition took place from humid-steppe biome or forest-steppe vegetation that was dominant during periods of soil formation (interglacial/interstadial) towards aridity-adapted semi-desert shrub species during glacial periods. In that case, n-alkane-derived grass signals would not be an indication of dry environmental conditions as widely supposed, but rather of a more humid situation.
At the moment we do prefer the third hypothesis as it seems to be likewise supported by gastropod analyses on the investigated LPS (Richter et al., in prep). The analysis of n-alkane biomarkers is still a promising tool to reconstruct paleoenvironmental changes. However, we think that it is very advisable to combine this analysis with the survey of further proxy information such as pollen compositions, land snail assemblages, or isotope measurements (e.g., 15N), in order to boost interpretation.