δ²H_n-alkane and δ¹⁸O_sugar biomarker proxies from leaves and topsoils of the Bale Mountains, Ethiopia, and implications for paleoclimate reconstructions

Abstract

The hydrogen isotopic composition of leaf wax–derived n-alkane (δ²H_n-alkane) and oxygen isotopic composition of hemicellulose–derived sugar (δ¹⁸O_sugar) biomarkers are valuable proxies for paleoclimate reconstructions. Here, we present a calibration study along the Bale Mountains in Ethiopia to evaluate how accurately and precisely the isotopic composition of precipitation is imprinted in these biomarkers. n-Alkanes and sugars were extracted from the leaf and topsoil samples and compound–specific δ²H_n-alkane and δ¹⁸O_sugar values were measured using a gas chromatograph–thermal conversion–isotope ratio mass spectrometer (GC–TC–IRMS). The weighted mean δ²H_n-alkane and δ¹⁸O_sugar values range from − 186 to − 89‰ and from + 27 to + 46‰, respectively. Degradation and root inputs did not appear to alter the isotopic composition of the biomarkers in the soil samples analyzed. Yet, the δ²H_n-alkane values show a statistically significant species dependence and δ¹⁸O_sugar yielded the same species–dependent trends. The reconstructed leaf water of Erica arborea and Erica trimera is ²H– and ¹⁸O–enriched by + 55 ± 5 and + 9 ± 1‰, respectively, compared to precipitation. By contrast, Festuca abyssinica reveals the most negative δ²H_n-alkane and least positive δ¹⁸O_sugar values. This can be attributed to “signal–dampening” caused by basal grass leaf growth. The intermediate values for Alchemilla haumannii and Helichrysum splendidum can be likely explained with plant physiological differences or microclimatic conditions affecting relative humidity (RH) and thus RH–dependent leaf water isotope enrichment. While the actual RH values range from 69 to 82% (x̄ = 80 ± 3.4%), the reconstructed RH values based on a recently suggested coupled δ²H_n-alkane –δ¹⁸O_sugar (paleo–) hygrometer approach yielded a mean of 78 ± 21%. Our findings corroborate (i) that vegetation changes, particularly in terms of grass versus non–grassy vegetation, need to be considered in paleoclimate studies based on δ²H_n-alkane and δ¹⁸O_sugar records and (ii) that the coupled δ²H_n-alkane –δ¹⁸O_sugar (paleo–) hygrometer approach holds great potential for deriving additional paleoclimatic information compared to single isotope approaches.

Introduction

The compound–specific hydrogen isotopic composition of leaf wax–derived n-alkanes (δ²H_n-alkane) and oxygen isotopic composition of hemicellulose–derived sugars (δ¹⁸O_sugar) serve as valuable proxies in (paleo)–climate and –environmental studies (e.g. Zech et al. 2013b, 2014a; Tuthorn et al. 2015; Hepp et al. 2017). The above mentioned biomarkers are preserved for a long period of time (Eglinton and Hamilton 1967; Glaser and Zech 2005; Eglinton and Eglinton 2008) and their application was tested in different paleoclimate archives, such as for instance in lake sediments (Zech et al. 2014b; Hepp et al. 2015), loess–paleosols (Zech et al. 2013b; Buggle and Zech 2015), and tree rings (Gessler et al. 2009). The isotopic composition of δ²H_n-alkane and δ¹⁸O_sugar from terrestrial vascular plants primarily reflects the isotopic composition of precipitation. In turn, the isotopic composition of δ²H_prec and δ¹⁸O_prec on the continents is mainly controlled by different site and climatic factors. The most prominent factors determining the spatial and temporal variation of δ²H_prec and δ¹⁸O_prec were attributed to ‘effects’ (Dansgaard 1964; Rozanski et al. 1993). For instance, the ‘temperature–effect’ describes that at lower/higher temperatures δ²H_prec and δ¹⁸O_prec values become more negative/positive particularly at higher latitudes. Particularly in many tropical regions, the ‘amount–effect’ describes that low/high amounts of precipitation typically coincide with more positive/negative δ²H_prec and δ¹⁸O_prec values. And finally, more negative δ²H_prec and δ¹⁸O_prec values are usually observed at higher elevation due to the ‘altitude–effect’. Recent publications emphasize, however, that other factors than δ²H_prec can exert strong control over δ²H_n-alkane, thus preventing δ²H_prec as well as e.g. paleoaltimetry reconstructions (Zech et al. 2015; Coffinet et al. 2017; Jaeschke et al. 2018).

Some studies reveal the absence of isotope fractionation during decomposition (Zech et al. 2012b) and water uptake by the roots (White et al. 1985; Ehleringer et al. 1991; Feakins and Sessions 2010). Nevertheless, caution must be used when using δ²H_n-alkane for paleoclimate reconstruction due to the effects of degradation and synthesis of leaf wax–derived n-alkanes by microorganisms (Brittingham et al. 2017; Zech et al. 2009, 2012a). Similarly, investigations on halophytic and xerophytic plants show evidence of hydrogen isotope fractionation during water uptake at the root–soil interface (Ellsworth and Williams 2007; Ladd and Sachs 2015). Apart from this, it is well known that leaf water is ²H– and ¹⁸O–enriched compared to plant source water due to evapotranspiration. This enrichment is primarily dependent on relative humidity (RH) and can challenge a robust δ²H_prec and δ¹⁸O_prec reconstruction (Zech et al. 2015). Moreover, the evaporative enrichment of leaf water affects the δ²H_n-alkane and δ¹⁸O_sugar values of monocot and dicot plants at different magnitudes. This latter effect is called “signal–dampening” and causes grasses to have generally more negative/less positive compound–specific stable hydrogen and oxygen isotope values than broad–leaf plant species (Hepp et al. 2019, 2020a). Hence, the interpretation of the isotopic composition of plant–derived biomarkers should consider (i) the isotopic composition of precipitation, (ii) RH– dependent evapotranspirative enrichment of leaf water, (iii) biosynthetic fractionation and signal– dampening (cf. section ‘Conceptual framework for interpreting the coupled δ²H_n-alkane–δ¹⁸O_sugar approach: reconstruction of leaf water, source water, d-excess, and relative humidity’).

Some years ago, Zech et al. (2013b) proposed a coupled δ²H_n-alkane –δ¹⁸O_sugar approach. This coupled approach aims at providing additional information to reconstruct paleoclimate and allows the reconstruction of δ²H and δ¹⁸O source water (δ²H_{source water} and δ¹⁸O_{source water}) and deuterium excess (d-excess, a proxy for RH). This approach was validated using climate transects across Argentina (Tuthorn et al. 2015) and South Africa (Strobel et al. 2020) as well as using a climate chamber experiment (Hepp et al. 2020b) and yielded highly significant correlations between actual and reconstructed RH and precipitation. The coupled approach was also successfully applied to eolian/loess–paleosol sediments and lacustrine sediments to reconstruct paleoclimate changes (Zech et al. 2013b; Hepp et al. 2017, 2019). The concept of using both δ²H and δ¹⁸O was independently developed for tree–ring cellulose to deduce RH (Voelker et al. 2014).

The main objective of the present study was to evaluate how accurately and precisely the isotopic signature of contemporary precipitation and RH are imprinted on compound–specific stable isotopes of leaves and topsoils of the Bale Mountains. More specifically, we aim at answering the following questions: (I) Are there systematic δ²H_n- alkane and δ¹⁸O_sugar differences between leaves, O–layers, and A_h–horizons? (II) Do n-alkane and sugar biomarkers in leaves and topsoils reflect δ²H_prec and δ¹⁸O_prec? (III) Are there systematic δ²H_n-alkane and δ¹⁸O_sugar differences among the locally dominant plant species? (IV) Does the application of the coupled δ²H_n-alkane –δ¹⁸O_sugar (paleo–) hygrometer approach provide additional information compared to the single isotope approaches?

Material and methods

Geography of the Bale Mountains

The Bale Mountains are located 400 km southeast of Addis Ababa, capital of Ethiopia, and belong to the Bale–Arsi Mountains massif (Fig. 1). The Bale Mountains National Park (BMNP) is situated between 39° 28′ to 39° 57′ longitude (E) and between 6° 29′ to 7° 10′ latitude (N), spanning a large altitudinal (1400–4377 m a.s.l.) gradient (Hillman 1986, 1988; Miehe and Miehe 1994; Tiercelin et al. 2008). The study area was described previously in detail by Lemma et al. (2020, 2019a, b). Briefly, the Bale Mountains are topographically divided into three major parts: The northern slopes (3000–3800 m. a.s.l.), the central Sanetti Plateau (3800–4377 m a.s.l), and the southern Harenna Escarpment (1400–3800 m. a.s.l.). The Sanetti Plateau is located in the center, on which the second highest peak of the country Mt. Tulu Dimtu (4377 m a.s.l.) perches (Friis 1986; Hillman 1986; Miehe and Miehe 1994). The northern slopes of the mountains are dissected by the Togona Valley, which descends towards the extensive Arsi Plateau and further down to the Great Rift Valley lowlands, which divide the country into two parts. The southern slope includes the steep Harenna Escarpment and goes down to the surrounding lowland at about 1400 m a.s.l. The vegetation shows an altitudinal zonation comprising the Afromontane forest (1450–2000 m a.s.l.), the upper montane forest dominated by Hagenia – Hypericum species (2000–3200 m a.s.l.), the Ericaceous belt (3200–3800 m a.s.l.), and the Afroalpine zone (3800–4377 m a.s.l.) being dominated by dwarf– shrubs (e.g., Helichrysum, Alchemilla), herbs, and grasses (mostly Festuca) (Hillman 1988; Woldu et al. 1989; Miehe and Miehe 1994).

Fig. 1

Map showing the location of Ethiopia in Africa, location of the Bale Mountains National Park (BMNP) in Ethiopia, and location of sampling site (light green star) and weather stations (yellow cross) along the southwest and northeast transect in the BMNP. (Color figure online)

The crust of the Bale Mountains is of volcanic origin, the soils which are developed from the basaltic, rhyolitic, and trachytic parent rock, can be generally characterized as silty loams of reddish–brown to black colour (Woldu et al. 1989). Nevertheless, the mountain's summit higher than 3000 m a.s.l were substantially flattened by repeated glaciations (for example between 20 and 12 cal kyr BP), the soils developed from such glacial rocks are usually shallow and gravelly (Hedberg 1964; Osmaston et al. 2005). Therefore, Andosols are the most ubiquitous soil type but also Cambisols and Leptosols are prevalent in some parts of the Bale Mountains. In the wetland and sedimentary basin, Gleysols and Histosols, and in rock shelters, Anthrosols are also a common soil type (Yimer et al. 2006; Billi 2015).

The Bale Mountains are historically vulnerable to extreme climate changes that depict Quaternary glacier cover with a spatial extent of about 180 km² (Umer et al. 2004; Osmaston et al. 2005) and anthropogenic impacts (Gil-Romera et al. 2019; Ossendorf et al. 2019). The chronology of lacustrine sediment from the lake Garba Guracha (a glacial lake at 3950 m a.s.l.) dated back to the Late Glacial (~ 16.5 cal kyr BP) illustrates the presence of regional climate events, the African Humid Period (AHP), that encompasses major dry fluctuations such as Younger Dryas (YD), 6.5 cal kyr BP and 4.2 cal kyr BP (Umer et al. 2007; Bittner et al. 2020).

Currently, the climate of the Bale Mountains is characterized by 4 months (November to February) of dry season and 8 months (March to October) of wet season (Kidane et al. 2012; Lemma et al. 2020). The seasonal migration of Intertropical Convergence Zone (ITCZ) determines the present climate and rainfall patterns of the Ethiopian Highlands (Tiercelin et al. 2008). The highest annual rainfall occurs in the southwestern part (1000–1500 mm per year) and the northern part of the mountains exhibits annual rainfall ranging between 800–1000 mm per year (Miehe and Miehe 1994; Tiercelin et al. 2008). Moisture during the dry season originates primarily from the Arabian Sea, whereas during the wet season it originates primarily from the Southern Indian Ocean (Lemma et al. 2020). The Afroalpine regions of the Bale Mountains are characterized by a large diurnal temperature range of 40 °C (− 15 to + 26 °C) (Hedberg 1964; Hillman 1988). Even if there is a wider variability of climate in the Bale Mountains, only few records are reported. The highest mean annual temperatures in the mountains peaks range between 6 and 12 °C. At Dinsho (headquarter of the BMNP, 3170 m a.s.l.) the mean annual temperature is 11.8 °C. The lowest mean annual temperatures range from 0.6 to 10 °C with frequent frost occurring in the high peak areas during the winter season (Tiercelin et al. 2008).

Sampling

For the present study, 25 leaf samples of locally dominant plant species were collected (Fig. 1) along a southwestern and a northeastern transect (ranging from 2550 to 4377 m a.s.l. and 3870 to 4134 m a.s.l., respectively). Samples comprise Erica trimera (n = 5), Erica arborea (n = 5), Alchemilla haumannii (n = 5), Festuca abyssinica (n = 6), Helichrysum splendidum (n = 2), Kniphofia foliosa (n = 1) and Lobelia rhynchopetalum (n = 1). Photos illustrating the investigated plant species were previously published by Mekonnen et al. (2019). Additionally, 15 organic surface layers (= O–layers, strongly humified plant residues) and 22 mineral topsoils (= A_h–horizons, 0–10 cm soil depth) that developed under the above–listed locally dominant vegetation were collected, resulting in 62 samples in total. All samples were air–dried in the Soil Store Laboratory of the National Herbarium, Department of Plant Biology and Biodiversity Management, Addis Ababa University. In the laboratory of the Soil Biogeochemistry Group, Martin Luther University of Halle–Wittenberg, the samples were sieved using a mesh size of 2 mm, finely ground, homogenized, and subjected to further biogeochemical analysis.

Additionally, from ten newly established weather stations, we recorded hourly RH, hourly temperature, and a total of 164 precipitation samples were collected along an altitudinal transect ranging from 1304 to 4375 m a.s.l. The δ²H_prec and δ¹⁸O_prec as well as mean annual RH and temperature records above 2550 m a.s.l. were used for the present study. For more details, information regarding the spatial and temporal ²H and ¹⁸O isotope variation of the contemporary precipitation in the Bale Mountains, the readers are referred to Lemma et al. (2020).

Compound–specific δ²H_n-alkane and δ¹⁸O_sugar analyses

Leaf wax–derived n-alkanes were extracted from 0.5 to 1 g of leaves, O–layers and A_h–horizons using Soxhlet extraction by adding 150 ml of dichloromethane (DCM) and methanol (MeOH) as solvents (9:1 ratio) for 24 h following a method modified after Zech and Glaser (2008). 50 µl of 5α-androstane were added to the total lipid extracts as an internal standard. Total lipid extracts were separated over aminopropyl pipette columns (Supelco, 45 μm). Three lipid fractions containing the n-alkanes, alcohols, and fatty acids were eluted successively by using 3 ml of hexane, DCM/MeOH (1:1), and diethyl ether and acetic acid (95:5), respectively.

The compound–specific stable hydrogen isotope analyses of the most abundant n-alkanes (C₂₉ and C₃₁) were determined using a Trace GC 2000 coupled to a Delta V Advantage IRMS via a thermal conversion reactor (GC IsoLink) and a ConFlo IV interface (GC–TC–IRMS, Delta V, Thermo Fisher Scientific, Bremen, Germany) operating in pyrolysis mode with reactor temperature of 1420 °C. The GC was equipped with a split/splitless injector and an Rtx–5 Sil MS column (30 m length, 0.25 mm inner dimeter, 0.5 μm film thickness). Each sample was analysed in triplicate and measurements of n-alkane standards (n-C₂₇, n-C₂₉, n-C₃₃) with known isotopic composition (Arndt Schimmelmann, Indiana) at various concentrations were embedded in–between the three sample batches. The ${\mathrm{H}}_{3}^{+}$ factor was determined before and after each sequence and was stable at 7.57 ± 0.29 (n = 6) during the measurement campaign.

The hemicellulose–derived sugars were extracted hydrolytically (105 °C, 4 h) from aliquots of leaves, O–layers, and A_h–horizons using 10 ml of 4 M trifluoroacetic acid (TFA) following the method described by Zech and Glaser (2009). The extracted monosaccharides were purified via XAD–7 and Dowex 50WX8 resin columns as proposed by Amelung et al. (1996). For quantification, one–fifth of the freeze–dried sugar samples were derivatized using N-methyl-pyrrolidone (NMP) and heated for 30 min at 75 °C. After cooling, the samples were subjected for the second derivatization using 400 μl bis(trimethylsilyl) trifluoroacetamide (BSTFA) and heated again for 5 min at 75 °C. For compound–specific δ¹⁸O_sugar analyses, the remaining four–fifth of the samples were derivatized using the methyl–boronic acid (MBA; 4 mg in 400 µl pyridine) derivatization method and heated for 1 h at 60 °C (Zech and Glaser 2009).

The compound–specific stable oxygen isotope values were determined using a Trace GC 2000 coupled to a Delta V Advantage IRMS via an ¹⁸O–pyrolysis reactor (GC IsoLink) and a ConFlo IV interface (GC–Py–IRMS, Delta V, Thermo Fisher Scientific, Bremen, Germany). The GC was equipped with a split/splitless injector and a DB– 5MS + DG column (60 m length, 0.250 mm inner dimeter, 0.25 μm film thickness). Samples were injected in splitless mode and analysed in triplicate. For calibration standard blocks of derivatised sugars (arabinose, fucose and xylose) of known isotopic composition at various concentrations were embedded in–between the six sample batches.

The measured δ²H_n-alkane and δ¹⁸O_sugar values were corrected for drift and amount–dependence, according to Zech and Glaser (2009). Out of 62 samples, five samples (four O–layers and one A_h–horizon below Erica) were rejected from further data evaluation and interpretation due to peak areas below detection limit. Similarly, the sugar biomarker fucose was excluded from data evaluation due to too low peak areas. Mean standard error for triplicate measurements of the remaining 57 samples are 0.8‰, 0.9‰, 3.1‰, 2.3‰ for arabinose, xylose, n-C_29, and n-C₃₁ alkanes, respectively. The standard errors for the sugar and n-alkane standards were 2‰ (n = 22) and 5‰ (n = 28), respectively. In the following, δ²H_n-alkane values refer to weighted mean values of n-C₂₉ and n-C_31, and δ¹⁸O_sugar values refer to weighted mean values of arabinose and xylose. The weighted mean for δ²H_n-alkane and δ¹⁸O_sugar values were calculated using the relative amounts of n-C₂₉ and n-C₃₁ as well as arabinose and xylose, respectively. The δ²H_n-alkane and δ¹⁸O_sugar values are reported in per mil (‰) according to the usual δ–notation relative to Vienna Standard Mean Ocean Water (V–SMOW) given in Eq. 1.

$$\updelta = \left( {{\text{R}}_{{{\text{sample}}}} /{\text{R}}_{{{\text{standard}}}} - 1} \right) \times 10^{3}$$

(1)

where R refers to the ratio of ¹⁸O/¹⁶O and ²H/¹H for the sample or standard (V–SMOW) materials.

Conceptual framework for interpreting the coupled δ²H_n-alkane–δ¹⁸O_sugar approach: reconstruction of leaf water, source water, d-excess, and relative humidity

The isotopic composition of leaf water (δ²H_{leaf water} and δ¹⁸O_{leaf water}) can be reconstructed from the isotopic composition of biomarkers (δ²H_n-alkane and δ¹⁸O_sugar) by subtracting biosynthetic fractionation factors (Zech et al. 2013b; Tuthorn et al. 2015; Hepp et al. 2017, 2019). Apart from assuming that the biosynthetic fractionation factors (ε_bio) are constant and not significantly variable amongst plants, we assume moreover that both biomarkers (sugars and n-alkanes) are primarily leaf–derived. We acknowledge that both assumptions are valid only in approximation. We apply ε_bio of − 160‰ (Sessions et al. 1999) to reconstruct δ²H_{leaf water} (Eq. 2) and + 27‰ (Sternberg 1989; Schmidt et al. 2001; Cernusak et al. 2003; Gessler et al. 2009) to reconstruct δ¹⁸O_{leaf water} (Eq. 3). These values were recently validated by Hepp et al. (2020b) using a dataset obtained from a well–controlled climate chamber experiment.

$${\updelta }^{{{18}}} {\text{O}}_{{\text{leaf water}}} = {{\left( {{\updelta }^{{{18}}} {\text{O}}_{{{\text{sugar}}}} {-}{\upvarepsilon }^{{{18}}}_{{{\text{bio}}}} } \right)} \mathord{\left/ {\vphantom {{\left( {{\updelta }^{{{18}}} {\text{O}}_{{{\text{sugar}}}} {-}{\upvarepsilon }^{{{18}}}_{{{\text{bio}}}} } \right)} {\left( {{{{1} + {\upvarepsilon }^{{{18}}}_{{{\text{bio}}}} } \mathord{\left/ {\vphantom {{{1} + {\upvarepsilon }^{{{18}}}_{{{\text{bio}}}} } {{1}000}}} \right. \kern-\nulldelimiterspace} {{1}000}}} \right)}}} \right. \kern-\nulldelimiterspace} {\left( {{{{1} + {\upvarepsilon }^{{{18}}}_{{{\text{bio}}}} } \mathord{\left/ {\vphantom {{{1} + {\upvarepsilon }^{{{18}}}_{{{\text{bio}}}} } {{1}000}}} \right. \kern-\nulldelimiterspace} {{1}000}}} \right)}}$$

(2)

$$\updelta^{{\text{2}}} {\text{H}}_{{{\text{leaf}}\;{\text{water}}}} = {{\left( {\updelta^{{\text{2}}} {\text{H}}_{{n{\text{-alkane}}}} - \upvarepsilon_{{{\text{bio}}}}^{2} } \right)} \mathord{\left/ {\vphantom {{\left( {\updelta^{{\text{2}}} {\text{H}}_{{n{\text{ - alkane}}}} - \upvarepsilon_{{{\text{bio}}}}^{2} } \right)} {\left( {{{1 + \upvarepsilon_{{{\text{bio}}}}^{2} } \mathord{\left/ {\vphantom {{1 + \upvarepsilon_{{{\text{bio}}}}^{2} } {1000}}} \right. \kern-\nulldelimiterspace} {1000}}} \right)}}} \right. \kern-\nulldelimiterspace} {\left( {{{1 + \upvarepsilon_{{{\text{bio}}}}^{2} } \mathord{\left/ {\vphantom {{1 + \upvarepsilon_{{{\text{bio}}}}^{2} } {1000}}} \right. \kern-\nulldelimiterspace} {1000}}} \right)}}$$

(3)

Once the isotopic composition of leaf water is calculated according to Eqs. 2 and 3, it is possible to reconstruct the isotopic composition of source water, d-excess and RH according to Eqs. (4), (5), and (6), respectively.

Leaf water becomes ²H– and ¹⁸O–enriched due to evapotranspiration via stomata and will reach isotope steady–state under constant environmental conditions. This process is accompanied by equilibrium and kinetic isotope effects. The isotopic composition of δ²H_{leaf water} and δ¹⁸O_{leaf water} then depends primarily on RH, the isotopic composition of source water and the isotopic composition of atmospheric water vapor (Dongmann et al. 1974; Flanagan et al. 1991; Roden and Ehleringer 1999). Once isotope steady–state is achieved, the isotope ²H– and ¹⁸O–enrichment of leaf water can be described using a Craig–Gordon model (Craig and Gordon 1965) adapted by Gat and Bowser (1991) and later by Zech et al. (2013b). The isotopic composition of leaf water at the evaporative site can be calculated using Eq. 4.

$${\updelta }_{{\text{leaf water}}} = {\updelta }_{{\text{source water}}} + \left( {{1} - {\text{RH}}} \right){\upvarepsilon }^{*} + \Delta {\upvarepsilon }$$

(4)

where δ_{leaf water} and δ_{source water} are the isotopic composition of leaf water and source water, respectively, RH is relative humidity, ε^* is equilibrium isotope enrichment ((1 − 1/α_L/V) $\text{*}$ 10³), and ∆ε is kinetic isotope enrichment [∆¹⁸ε = ${\text{C}}_{{\text{k}}}^{{18}} (1 - {\text{RH}}$); ∆²ε = ${\text{C}}_{{\text{k}}}^{2} (1 - {\text{RH}}$)]. ${\text{C}}_{\text{k}}^{18}$ and ${\text{C}}_{\text{k}}^{2}$ are kinetic isotope enrichment constants for ¹⁸O and ²H, respectively.

In a two–dimensional δ²H–δ¹⁸O diagram, the isotope values of precipitation plot along the GMWL (Craig 1961; Dansgaard 1964), whereas due to transpiration leaf water becomes ²H– and ¹⁸O–enriched compared to source water (precipitation) (Farquhar et al. 2007; Feakins and Sessions 2010) and the isotope values of leaf water plot along the evaporation line (Zech et al. 2013b; Tuthorn et al. 2015). The distance of the reconstructed leaf water to the GMWL can be described as d-excess and serves as a proxy for relative humidity (Fig. 2). d-excess is a second– order stable isotope parameter measured in meteoric water defined as d-excess = δ²H – 8*δ¹⁸O. Therefore, d-excess of leaf water can be derived using Eq. 5.

$$d_{{{\text{leaf}}\;{\text{water}}}} = d_{{{\text{source}}\;{\text{water}}}} + \left( {1 - {\text{RH}}} \right)\left( {{\upvarepsilon }_{2}^{*} - 8 * {\upvarepsilon }_{18}^{*} + {\text{C}}_{{\text{k}}}^{2} - 8*{\text{C}}_{{\text{k}}}^{18} } \right)$$

(5)

Fig. 2

δ¹⁸O versus δ²H diagram illustrating the conceptual framework of the coupled (paleo–) hygrometer approach after Zech et al. (2013b). Data are plotted for measured precipitation (blue solid triangle), measured δ²H_n-alkane, and δ¹⁸O_sugar biomarkers (black solid circle), reconstructed leaf water (green solid rectangle), reconstructed source water (light blue solid rectangle), measured Garba Guracha lake water (brown solid triangle), global meteoric water line, GMWL (black solid line) and local evaporation line, LEL (red solid line). (Color figure online)

where d_{leaf water} and d_{source water} are d-excess of leaf water and source water, respectively and the equilibrium ($\upvarepsilon _{2}^{*} \;{\text{and}}\;\upvarepsilon _{{18}}^{*} {\text{) and kinetic (C}}_{{\text{k}}}^{2} {\text{ and C}}_{{\text{k}}}^{{18}}$) isotope fractionation of the two isotopes. Equation 5 illustrates that d-excess depends primarily on RH. Therefore, RH normalized to the temperature of leaf water can be estimated using Eq. 6.

$${\text{RH}} \approx 1 - \frac{{\Delta d}}{{\left( {{\upvarepsilon}_{2}^{*} - 8*{\upvarepsilon}_{{18}}^{*} + {\text{C}}_{{\text{k}}}^{2} - 8*{\text{C}}_{{\text{k}}}^{{18}} } \right)}}$$

(6)

where Δd is the difference between d-excess of leaf water and d-excess of source water. Equilibrium isotope fractionation factors (${\varepsilon }_{2}^{*}$ and ${\varepsilon }_{18}^{*}$) as a function of temperature are calculated according to empirical equations of Horita and Wesolowski (1994). According to Merlivat (1978), the maximum values of kinetic isotope fractionation factors (${\mathrm{C}}_{\mathrm{k}}^{2}$ and ${\mathrm{C}}_{\mathrm{k}}^{18}$) during molecular diffusion of water via a stagnant air for ²H and ¹⁸O are 25.1‰ and 28.5‰, respectively. However, we are aware that the kinetic isotope fractionation factors are dependent on aerodynamic conditions (liquid–vapor interface, diffusive, and turbulent layer) characterising a given evaporation process.

The values of δ²H_{source water} and δ¹⁸O_{source water} can be estimated when the slope of the evaporation line for leaf water is known. The slope of the evaporation line can be derived from Eq. 7.

$${\text{S}}_{{{\text{LEL}}}} { = }\frac{{{\updelta }_{{\text{e}}}^{{2}} - {\updelta }_{{\text{s}}}^{{2}} }}{{{\updelta }_{{\text{e}}}^{{{18}}} - {\updelta }_{{\text{s}}}^{{{18}}} }} \approx \frac{{{\upvarepsilon }_{{2}}^{*} { + }\,{\text{C}}_{{\text{k}}}^{{2}} }}{{{\upvarepsilon }_{{{18}}}^{*} { + }\,{\text{C}}_{{\text{k}}}^{{{18}}} }}$$

(7)

In this equation, we assumed that the isotopic composition of source water and atmospheric water vapor reached equilibrium. Thus, the slope of the evaporation line depends on equilibrium (${\varepsilon }_{2}^{*}$ and ${\varepsilon }_{18}^{*}$) and kinetic (${\text{C}}_{\text{k}}^{18}$ and ${\text{C}}_{\text{k}}^{2}$) isotope fractionation factors of the two isotopes (Gat and Bowser 1991). In the present study, the slope of the local evaporation line (LEL) ranges between 2.86 and 3.07 (x̄ = 2.97 ± 0.11) for the temperature ranging between 3.2 and 12.7 °C. Here in this manuscript, ± sign refer to standard error values. Such low slope for evaporating leaf water is in agreement with both field observations and laboratory experiments (Allison et al. 1985; Walker and Brunel 1990; Gat et al. 2007).

The net isotope difference (also called apparent fractionation, ɛ_app) between δ²H_n-alkane and δ²H_prec as well as δ¹⁸O_sugar and δ¹⁸O_prec is calculated using Eq. 8.

$${\varepsilon }_{\mathrm{app}}=1000\times \left(\begin{array}{c}\frac{{\delta }_{a }+1000}{{\delta }_{b}+1000}-1\end{array}\right)$$

(8)

where $a$ is the isotopic composition of a biomarker (δ²H_n-alkane or δ¹⁸O_sugar) and $b$ is the weighted mean annual isotopic composition of precipitation (δ²H_prec or δ¹⁸O_prec).

Results and discussion

Comparison of δ²H_n-alkane and δ¹⁸O_sugar results from leaves, O–layers, and A_h–horizons

The δ²H values of n-C₂₉ and n-C₃₁ alkanes as well as the δ¹⁸O values of arabinose and xylose correlate highly significant with each other (r = 0.9, p < 0.0001 and r = 0.8, p < 0.0001, respectively, ESM_1 and 2 of the supplementary files). In the following, we therefore use and refer to δ²H_n-alkane values for weighted mean values of n-C₂₉ and n-C₃₁ as well as to δ¹⁸O_sugar values for weighted mean values of arabinose and xylose. Weighted mean δ²H_n-alkane and δ¹⁸O_sugar values range from − 186 to − 89‰ and from + 27 to + 46‰, respectively (Fig. 3). Particularly, the topsoil (O–layer as well as A_h–horizon) samples in the Bale Mountains yielded a δ²H_n-alkane values range between − 157 and − 113‰ (x̄ =  − 136‰), which is in good agreement with findings from the highest Eastern African Mt. Kilimanjaro (Peterse et al. 2009; Zech et al. 2015; Hepp et al. 2017), Mt. Rungwe and Mt. Kenya (Coffinet et al. 2017).

Fig. 3

Altitudinal δ²H (left) and δ¹⁸O (right) gradients of n-alkanes (weighted mean of n-C₂₉ and n-C₃₁, solid circles) and sugars (weighted mean of arabinose and xylose, empty circles), respectively, reconstructed leaf water (squares) of all the investigated samples as well as measured weighted mean annual precipitation (triangles)

The δ²H_n-alkane and δ¹⁸O_sugar values of leaves, O–layers, and A_h–horizons do not depict statistically significant differences (p = 0.7 and p = 0.4, respectively) among each other (Fig. 4a and b). Hence, our data provide no evidence for degradation effects affecting the isotopic composition of the biomarkers. Similarly, there is no evidence for root and soil microbial–derived n-alkanes and sugars affecting the compound–specific hydrogen and oxygen isotope signals. This is remarkable because our previous chemotaxonomy studies reveal that the characteristic plant biomarker patterns of long-chain n-alkanes and sugars are strongly altered in the respective O–layers and A_h–horizons due to degradation and/or underground input by roots (Lemma et al. 2019b; Mekonnen et al. 2019).

Fig. 4

Comparison of δ²H_n-alkane (a) and δ¹⁸O_sugar (b) among leaves, O–layers and A_h–horizons. The notched box plots indicate the median (solid lines between the boxes), interquartile range (IQR) with upper (75%) and lower (25%) quartiles. The notches display the confidence interval around the median within ± 1.57*IQR/$\sqrt{n}$

Concerning δ¹⁸O_sugar, Zech et al. (2012b) reported the absence of isotope fractionation during decomposition based on a litter decay study. By contrast, based on the same litter decay sample set, Zech et al. (2011) concluded that the built-up of a microbial n-alkane pool can change the δ²H_n-alkane values of soils compared to litter. The same conclusion was drawn by Tu et al. (2011) based on a δ¹³C n-alkane litter decay study. A possible explanation for this discrepancy could be that δ²H_n-alkane and δ¹⁸O_sugar variability in our dataset are very large within the leaf, O– layer, and A_h–horizon sub-dataset (Fig. 4a and b). Indeed, albeit not significant, the median values of O–layers reveal both the least negative δ²H_n-alkane and the most positive δ¹⁸O_sugar. It is noteworthy that the concentration of n-alkanes in roots is much lower than in leaves and the contamination of leaf wax–derived n-alkanes in sediment by the roots is negligible (Zech et al. 2012a, 2013a), whereas a partial contribution of especially root–derived sugars to O–layers and A_h–horizons is very likely. Still, the mean δ²H_n-alkane and δ¹⁸O_sugar values of leaves and A_h–horizons are almost identical (− 134 versus − 135‰ and 35 versus 36‰, respectively). Hence, our dataset does not provide evidence for degradation effects and root input affecting δ²H_n-alkane and δ¹⁸O_sugar values of topsoils.

Comparison of δ²H_n-alkane and δ¹⁸O_sugar results with δ²H_prec or δ¹⁸O_prec results

The δ²H_n-alkane values correlate significantly with the weighted mean annual isotopic composition of local meteoric water, δ²H_prec (r = 0.3, p = 0.04, n = 53, ESM_3 of the supplementary files). However, it does not correlate in a 1:1 relationship (δ²H_prec = 0.05*δ²H_n-alkane – 1.03) and this is the first important indication that n-alkanes do not directly reflect δ²H_prec. Instead, δ²H_n-alkane reflects the isotope signal of local precipitation modified by leaf water enrichment as emphasized by Zech et al. (2015). The isotopic composition of precipitation in the Bale Mountains is subjected to altitude, amount, and seasonal effects. The most positive δ²H_prec and δ¹⁸O_prec, as well as the more negative δ²H_prec and δ¹⁸O_prec values, were recorded in the lowermost and highest weather stations of the Bale Mountains, respectively (Lemma et al. 2020). However, the overall significant trend of the altitudinal effect observed in modern precipitation is not well reflected in the δ²H_n-alkane and δ¹⁸O_sugar (Fig. 3). Our findings are hence in agreement with δ²H_n-alkane results from Mt. Kilimanjaro (Peterse et al. 2009; Zech et al. 2015), Mt. Rungwe (Coffinet et al. 2017) and the southwest Ethiopian highlands (Jaeschke et al. 2018). There is also no significant correlation between δ¹⁸O_sugar and δ¹⁸O_prec in our dataset. This might indicate that the isotope signal of δ¹⁸O_prec is strongly altered prior to the biosynthesis of the sugar biomarkers. Thus, soil water evaporation, leaf water evapotranspiration and seasonality effects are the most likely responsible factors. A recent publication by Strobel et al. (2020) argued that plant–related and/or environmental factors strongly bias the δ¹⁸O_sugar signal compared to δ¹⁸O_prec and points especially to the significant impact of evapotranspirative soil and leaf water enrichment. Similarly, a climate transect study revealed that δ¹⁸O_{hemicellulose} does not reflect δ¹⁸O_prec along an Argentinian transect, either (Tuthorn et al. 2014). Overall it can be summarized that our δ²H_n-alkane and δ¹⁸O_sugar values do not directly reflect the isotopic composition of precipitation in the Bale Mountains.

The apparent isotope fractionation (ɛ_app) for each investigated plant species (only leaf samples) ranges for hydrogen (ɛ_{app 2H}) between − 180 to − 86‰ and for oxygen (ɛ_{app 18O}) between + 25 to + 56‰. The ɛ_app factors fundamentally depend on three processes (Fig. 5, after Strobel et al. 2020). Accordingly, the main processes influencing the imprint of precipitation isotope on biomarkers are soil–water evaporation, leaf–water transpiration, and biosynthetic fractionation. The mean ɛ_{app 2H} and ɛ_{app 18O} values of the investigated plant species in the Bale Mountains are − 128 ± 6‰ and + 40 ± 2‰, respectively. This is in a good agreement and within the range of results reported from a winter rainfall zone transect study in South Africa (Strobel et al. 2020).

Fig. 5

Conceptual model demonstrating the hydrogen–isotope and oxygen–isotope relationships, between precipitation, leaf water, and leaf wax–derived n-alkanes as well as hemicellulose–derived sugars (after Strobel et al. 2020). ɛ_{app 2H}: apparent hydrogen-isotope fractionation, ɛ_{app 18O}: apparent oxygen-isotope fractionation, ε¹⁸_bio: biosynthetic oxygen–isotope fractionation, ε²_bio: biosynthetic hydrogen-isotope fractionation

In the present study, there is no significant correlation between apparent isotope fractionation and mean day time temperature (ɛ_{app 18O}, r = 0.1, p = 0.5 and ɛ_{app 2H}, r = 0.3, p = 0.1). Thus, the apparent isotope fractionation is not influenced by temperature–related evapotranspirative enrichment. This is in agreement with the results of Strobel et al. (2020) who found that neither ɛ_{app 2H} nor ɛ_{app 18O} is significantly dependent on mean annual temperature (MAT).

Interspecies comparison of δ²H_n-alkane and δ¹⁸O_sugar results from dominant plant species in the Bale Mountains

Combining all leaf, O–layer and A_h–horizon data, the comparison of the δ²H_n-alkane values between the plant species under study depicts significant differences (Fig. 6a). More specifically, Erica arborea and Erica trimera yielded the least negative median δ²H_n-alkane values, whereas Festuca abyssinica yielded the most negative median δ²H_n- alkane value (Fig. 6a). Albeit the δ¹⁸O_sugar values show no significant difference amongst the dominant plant species (Fig. 6b), they reveal the same trends as the δ²H_n-alkane results. Particularly, E. arborea and E. trimera depicted the most positive median δ¹⁸O_sugar values, whereas F. abyssinica depicted the least positive median δ¹⁸O_sugar value (Fig. 6b). The mean δ²H_n-alkane and δ¹⁸O_sugar values of E. arborea and E. trimera are − 114 and − 109‰, and + 37 and + 36‰, respectively. These very close mean δ²H_n-alkane and δ¹⁸O_sugar values of the two Erica species likely reflect the monophyletic nature (Guo et al. 2014).

Fig. 6

Variations of δ²H_n-alkane (a) and δ¹⁸O_sugar (b) between samples (leaf, O–layer and A_h–horizon) of the dominant plant species. The notched box plots indicate the median (solid lines between the boxes), interquartile range (IQR) with upper (75%), and lower (25%) quartiles and outliers (empty circle). The notches display the confidence interval around the median within ± 1.57*IQR/$\sqrt{n}$

One might expect that the above described plant–specific differences can be explained with the altitude–dependent different occurrence of the investigated plant species and altitude–dependent δ²H_prec and δ¹⁸O_prec variability. Indeed, the δ²H_prec and δ¹⁸O_prec values in the Bale Mountains exhibit spatial variance along the altitudinal gradient with isotope lapse rates of − 0.52‰*100 m⁻¹ and − 0.11‰*100 m⁻¹, respectively (Lemma et al. 2020). This corresponds roughly to a range of 5 and 1‰ for δ²H_prec and δ¹⁸O_prec, respectively, along the investigated altitudinal transect (3226 and 4375 m a.s.l.). These ranges are, however, too small to explain the observed large δ²H_n-alkane and δ¹⁸O_sugar variability along the transect (Fig. 3). Strikingly, the median δ²H_n-alkane and δ¹⁸O_sugar values of F. abyssinica are the lowest compared to the other investigated plant species. This can be attributed to physiological and biochemical differences and the fact that F. abyssinica (grass) is a monocot, whereas the other plant species are dicots. The leaves of monocots and dicots differ in structure, location and timing of lipid synthesis (Sachse et al. 2012). The growth of grasses is associated with intercalary meristems. They occur at the base of the grass leaf blade, where leaf water is not fully ²H– and ¹⁸O–enriched. Thus, newly synthesized lipids and sugars at the base of grass leaves do not incorporate the full leaf water enrichment signal, which is called ²H_n-alkane and ¹⁸O_sugar “signal–damping or dampening effect” (Hepp et al. 2019, 2020a). On the other hand, the median δ²H_n-alkane and δ¹⁸O_sugar values of E. arborea and E. trimera are the highest in comparison with other species and show strong leaf water enrichment. The intermediate δ²H_n-alkane and δ¹⁸O_sugar values for A. haumannii and H. splendidum might be attributed to plant physiological and biochemical differences or microclimatic conditions affecting RH values and thus leaf water enrichment as compared to Erica and F. abyssinica.

Furthermore, there is a significant difference in ɛ_{app 2H} between the dominant plant species in the Bale Mountains. Unlike other species, E. arborea (− 94‰) and E. trimera (− 92‰) have the least negative and F. abyssinica (− 161‰) has the most negative ɛ_{app 2H}. It is known that C3 graminoid (Sachse et al. 2012) and C3 grasses (Gamarra et al. 2016) exhibited the most negative ɛ_{app 2H} compared to shrubs and trees. Such species–dependent variation of ɛ_{app 2H} can be attributed to intrinsic factors (e.g., plant physiological, phenological, photosynthesis pathway and biosynthesis–related), plant growth form, place and timing of leaf–wax synthesis, and stomatal nature (Liu and Yang 2008; Sachse et al. 2012; Gao et al. 2015; Griepentrog et al. 2019; Strobel et al. 2020). Therefore, caution must be given in terms of vegetation change when interpreting δ²H_n-alkane for paleoclimate reconstruction. There is no significant difference in ɛ_{app 18O} among dominant plant species in the Bale Mountains and H. splendidum shows the least positive ɛ_{app 18O} value (26‰). This finding might demonstrate that ɛ_{app 18O} is not strongly sensitive to vegetation type rather climate–related evapotranspirative enrichment. Previous studies of Tuthorn et al. (2014) and Strobel et al. (2020) described δ¹⁸O_sugar and ɛ_{app 18O} are significantly influenced by evapotranspiration and aridity index.

Coupled δ²H_n-alkane and δ¹⁸O_sugar biomarker results

Reconstructed δ²H_{leaf water} and δ¹⁸O_{leaf water} as well as δ²H_{source water} and δ¹⁸O_{source water}

The basic principle of the coupled δ²H_n-alkane –δ¹⁸O_sugar approach is based on the assumption that long–chain n-alkanes and pentose sugars are leaf–derived and reflect the isotopic composition of leaf water (Zech et al. 2013b, 2015; Tuthorn et al. 2015; Hepp et al. 2017, 2019). The isotopic composition of reconstructed leaf water is determined by the isotopic composition of local precipitation and evaporative leaf water enrichment. However, soil–water enrichment cannot be excluded but is considered to be negligible. Therefore, the reconstructed δ²H_{leaf water} and δ¹⁸O_{leaf water} allow us to reconstruct δ²H_{source water} and δ¹⁸O_{source water} as well as d-excess and RH. These valuable proxies and additional information can support a better understanding of the paleoclimate history, which cannot be addressed by a single isotope study. The reconstructed δ²H_{leaf water} and δ¹⁸O_{leaf water} values range between − 31 to 84‰ (x̄ = 31 ± 4‰) and 0.1 to 18‰ (x̄ = 8 ± 0.6‰), respectively (Fig. 3). Most of the reconstructed leaf water data plot below the GMWL (Fig. 2). The reconstructed leaf water data plotting below the GMWL is a typical feature of evaporation loss.

By applying a biosynthetic fractionation factor of − 160‰, the reconstructed δ²H_{leaf water} values highlight that the n-alkanes of F. abyssinica (Fig. 6a) reflect quite accurately and precisely the weighted mean annual δ²H_prec (− 2.7 ± 2‰) of the Bale Mountains (Lemma et al. 2020). However, it is known that the effect of evaporative leaf water isotope enrichment of grasses is negligible (Helliker and Ehleringer 2002; McInerney et al. 2011). By contrast, the reconstructed δ²H_{leaf water} values of E. arborea and E. trimera (Fig. 6a) do not reflect accurately and precisely the weighted mean annual δ²H_prec and the reconstructed δ²H_{leaf water} is higher compared to δ²H_prec by up to + 55 ± 5‰ (Fig. 6a). One well known and widely accepted factor responsible for this finding is leaf water enrichment. For instance, Zech et al. (2015) emphasized that δ²H_n-alkane data from a climate transect along the southern slopes of Mt. Kilimanjaro do not reflect δ²H_prec either and explained this with a strongly variable degree of leaf water enrichment along the altitudinal transect. Hence, our datasets clearly indicate that unlike Erica (dicots), the reconstructed δ²H_{leaf water} of F. abyssinica (monocot) does not fully expose to evaporative leaf water enrichment due to signal–dampening.

Similarly, considering a biosynthetic fractionation factor of + 27‰ (Cernusak et al. 2003), the reconstructed δ¹⁸O_{leaf water} values of all investigated plant species (Fig. 6b) do not reflect the isotopic composition of precipitation, δ¹⁸O_prec (− 3.3 ± 0.4‰) in the Bale Mountains (Lemma et al. 2020). More specifically, the reconstructed δ¹⁸O_{leaf water} values of E. arborea and E. trimera are higher compared to δ¹⁸O_prec by up to + 9 ± 1‰ (Fig. 6b). Indeed, yet it is not significant, the reconstructed δ¹⁸O_{leaf water} is also species–dependent. Tree ring research has shown that cellulose does not reflect the full leaf water δ¹⁸O enrichment signal (Gessler et al. 2013). Additionally, results from climate chamber experiments indicate that δ¹⁸O_{leaf water} values, reconstructed from sugar biomarker, are more positive compared to plant source water by up to + 25‰ (Zech et al. 2014a).

Moreover, the coupled δ²H_n-alkane –δ¹⁸O_sugar approach allows us to reconstruct the isotopic composition of source water as intersect of the local evaporation line (LEL) with the GMWL (Fig. 2). The reconstructed δ²H_{source water} and δ¹⁸O_{source water} values range from − 84 to 110‰ (x̄ = 8.6 ± 7‰) and − 12 to 12‰ (x̄ =  − 0.7 ± 0.8‰) (Figs. 2 and 7), respectively. Unlike δ²H_prec and δ¹⁸O_prec, the reconstructed δ²H_{source water} and δ¹⁸O_{source water} shows a large scattering (Fig. 7). The mean reconstructed δ²H_{source water} and δ¹⁸O_{source water} values reveal slight isotope enrichment as compared with the modern–day isotopic composition of δ²H_prec (− 2.7 ± 2‰) and δ¹⁸O_prec (− 3.3 ± 0.4‰) in the Bale Mountains, respectively (Lemma et al. 2020). Despite the large scattering of the reconstructed source water values this finding demonstrates that the coupled δ²H_n-alkane –δ¹⁸O_sugar approach allows to reconstruct the isotopic composition of source water quite accurately.

Fig. 7

The scatter plots to the left show the correlations between δ¹⁸O_{source water} (reconstructed) and δ¹⁸O_prec (measured) as well as δ²H_{source water} (reconstructed) and δ²H_prec (measured), respectively. The linear regression lines (black line), correlation coefficients (r), 95% confidence intervals (grey area) and significance values (p) as well as the 1:1 regression lines (grey lines) are provided for each diagram. The box plots to the right illustrate the offset of δ¹⁸O_{source water} (reconstructed) to δ¹⁸O_prec (measured) and δ²H_{source water} (reconstructed) to δ²H_prec (measured), respectively for the dominant plant species. The plots indicate the median (solid line between the boxes), interquartile range (IQR) with upper (75%) and lower (25%) quartiles and outliers (circles)

The correlation and offsets between the reconstructed source water and respective δ¹⁸O_prec as well as δ²H_prec are shown in Fig. 7. The offsets between the reconstructed source water and respective δ¹⁸O_prec and δ²H_prec show a highly significant difference (p = 0.001, n = 53) among the investigated plant species (box plots of Fig. 7). The median offset of F. abyssinica (δ¹⁸O =  − 3‰; δ²H =  − 33‰) and H. splendidum (δ¹⁸O =  − 3‰; δ²H =  − 36‰) are very close to each other and E. trimera (δ¹⁸O = 8‰; δ²H = 48‰) exhibits a markedly larger median offset than the other species (Fig. 7).

Deuterium excess, reconstructed relative humidity and its paleoclimate implication

The distance between the reconstructed leaf water and GMWL can be described as d-excess, which is the intercept of the meteoric water line (Dansgaard 1964), and serves as a proxy for RH (Hepp et al. 2017, 2019). The reconstructed d-excess values calculated using Eq. (5) range between − 105 and + 56‰. The most positive reconstructed d-excess values are likely caused by an overestimation of the δ²H and/or an underestimation of the δ¹⁸O values. Similarly, the most negative reconstructed d-excess values are likely caused by an underestimation of the δ²H and/or an overestimation of the δ¹⁸O values. The actual RH ranges between 69 and 82% (x̄ = 80 ± 3.4%) (Fig. 8). The reconstructed RH calculated using Eq. (6) reveals a much wider range of 40 to 121% (x̄ = 78 ± 21%), with eight anomalous values above 100% (Fig. 8). These anomalies can be explained by several uncertainties related to Δd, ε^*, and C_k values used in Eq. (6). The uncertainties of Δd mainly arise from analytical errors associated with the individual δ²H_n-alkane and δ¹⁸O_sugar measurement as well as uncertainties of the ε_bio. In approximation, the coupled approach of Zech et al. (2013b) suggests temperature–independent constant ε_bio values. However, concerning e.g. δ¹⁸O, this value is very likely an underestimation. Slightly higher ε_bio values for ¹⁸O and ²H of + 2‰ and + 10‰ would increase the reconstructed RH by + 8% and 6%, respectively (Hepp et al. 2017).

Fig. 8

The scatter plot to the left shows the correlation between reconstructed RH and actual RH. A 1:1 regression line (grey line) are provided. The box plots to the right illustrate the offset of reconstructed RH to actual RH for dominant plant species. The plots indicate the median (solid line between the boxes), interquartile range (IQR) with upper (75%) and lower (25%) quartiles

The actual mean annual day time RH measured along the transect is 80 ± 3.4%, which is + 2 ± 18% higher than the reconstructed mean RH. The reconstructed RH is humidity normalized to leaf temperature. Usually, leaf temperature is higher than the actual ambient temperature (Gates 1980). Thus, the difference between actual and reconstructed RH might be associated with leaf–air temperature differences (Hepp et al. 2017; Zech et al. 2013b). Additionally, our approach does not consider evaporative enrichment of soil water (Zech et al. 2013b; Hepp et al. 2015; Tuthorn et al. 2015). Moreover, most leaf–derived biomarkers are biosynthesized at an early stage of leaf ontogeny (Gershenzon et al. 2000; Tipple et al. 2013) and may not reflect the mean annual isotopic composition of precipitation, but rather the isotopic composition of precipitation during the growing season.

The reconstructed RH correlates highly significant and negatively with δ¹⁸O_{leaf water} (r =  − 0.7, p < 0.0001, n = 53, ESM_4 of the supplementary files) and positively with δ²H_{leaf water} (r = 0.6, p < 0.0001, n = 53, ESM_5 of the supplementary files). This result might demonstrate that signal changes in δ¹⁸O_{leaf water} due to RH are dissociated from signal changes in δ¹⁸O_prec. The negative correlation between the reconstructed RH and δ¹⁸O_{leaf water}, apparent in the present study, is also reported previously by Zech et al. (2013b). Thus, due to evaporation, the lower the reconstructed RH is the higher δ¹⁸O_{leaf water}. Moreover, the reconstructed RH values correlate highly significantly and positively with reconstructed δ²H_{source water} and δ¹⁸O_{source water} (r = 0.9, p < 0.0001, n = 53, ESM_6 and 7 of the supplementary files). A similar finding was reported by Hepp et al. (2017) applying the coupled δ²H_n- alkane –δ¹⁸O_sugar approach in the Mt. Kilimanjaro region.

Conclusion

The compound–specific stable hydrogen and oxygen isotopic composition of biomarkers (n-alkanes and sugars) analysed from leaves, O–layers, and A_h–horizons along an altitudinal transect in the Bale Mountains were evaluated. There are no systematic δ²H_n-alkane and δ¹⁸O_sugar trends and significant differences between leaves and corresponding O–layers, and A_h–horizons, respectively. Hence, there is no evidence from our dataset for degradation and underground root input affecting δ²H_n-alkane and δ¹⁸O_sugar values of topsoils. The dominant plant species in the Bale Mountains are characterized by statistically significant species–dependent δ²H_n-alkane values and, albeit not significant, also δ¹⁸O_sugar values. Substantial species–specific apparent isotope fractionation highlights the importance of investigating apparent as well as biosynthetic isotope fractionation factor of the representative flora at a given geographical region. As a broader implication vegetation changes need to be considered when aiming at reconstructing δ²H_prec from δ²H_n-alkanes. The coupled δ²H_n-alkane –δ¹⁸O_sugar approach allows reconstructing δ²H and δ¹⁸O of leaf water and plant source water. The latter show large scattering and on average 11.3‰ and 2.6‰ more positive δ²H and δ¹⁸O values, respectively, compared to precipitation. The reconstructed d-excess of leaf water enables us to reconstruct RH. The actual mean annual day time RH measured in the Bale Mountains is + 2 ± 18% higher than the reconstructed mean RH. The coupled δ²H_n-alkane –δ¹⁸O_sugar (paleo–) hygrometer approach as recently validated also by Hepp et al. (2020b) therefore holds great potential for deriving additional paleoclimatic information compared to single isotope approaches.