Controls on leaf water hydrogen and oxygen isotopes: a local investigation across seasons and altitude

The stable oxygen (δ¹⁸O_leaf) and hydrogen (δ²H_leaf) isotopes of leaf water act as a bridge that connects the hydroclimate to plant-derived organic matter. However, it remains unclear whether the source water (i.e., twig water, soil water, and precipitation) or meteorological parameters (i.e., temperature, relative humidity, and precipitation) are the dominant controls on δ¹⁸O_leaf and δ²H_leaf. Here, we reported a seasonal analysis of δ¹⁸O_leaf and δ²H_leaf together with isotopes from potential source waters and meteorological parameters along an elevation transect on the Chinese Loess Plateau. We found that δ²H_leaf values were more closely correlated with source water isotopes than δ¹⁸O_leaf values, whereas δ¹⁸O_leaf and δ²H_leaf values were similarly correlated with meteorological parameters along the elevation transect. Dual-isotope analysis showed that the δ¹⁸O_leaf and δ²H_leaf values were closely associated because of their similar altitudinal and seasonal responses, generating a well-defined isotope line relative to the local meteoric water line (LMWL). We also compared the measured δ¹⁸O_leaf and δ²H_leaf values with values predicted by the Craig–Gordon model and found no significant differences between them. We demonstrate that the first-order control on δ¹⁸O_leaf and δ²H_leaf values was the source water, and the second-order control was the enrichment associated with biochemical and environmental factors on the Loess Plateau.

1 Introduction

The stable isotope compositions of oxygen and hydrogen (δ¹⁸O and δ²H, respectively) are increasingly being used as powerful tracers to follow the path of water from its input as precipitation, movement through the soil, and ultimately to its release as soil evaporation and leaf transpiration (Penna and Van Meerveld, 2019). Leaf water transpiration plays a key role in regulating the water balance at scales ranging from catchment to global. Terrestrial plants can enrich heavier isotopes (²H and ¹⁸O) in leaf water via evaporative fractionation through the stoma (Helliker and Ehleringer, 2000; Liu et al., 2015; Cernusak et al., 2016), which is highly dependent on atmospheric conditions (e.g., temperature and relative humidity) and biophysiological processes (Farquhar et al., 2007; Kahmen et al., 2011; Cernusak et al., 2016). Subsequently, the isotopic signals from leaf water are integrated into plant organic matter, such as cellulose (e.g., Barbour, 2007; Lehman et al., 2017) and leaf wax (Liu et al., 2016, 2021b), as powerful proxies used for paleoclimate reconstruction (Pagani et al., 2006; Schefuß et al., 2011; Hepp et al., 2020). However, although leaf water isotopes are the fundamental parameters in ecohydrology and organic biosynthesis, an adequate understanding of controls of leaf water isotopes and the role of source water and hydroclimate in determining leaf water isotopes is still lacking.

δ¹⁸O_leaf and δ²H_leaf values are influenced first by a plant's source water (mainly water taken up by roots from the soil; Cernusak et al., 2016; Barbour et al., 2017; Munksgaard et al., 2017; Liu et al., 2022) and second by the enrichment associated with transpiration (Munksgaard et al., 2017). Soil water for terrestrial plants generally originates from local precipitation, and precipitation isotopes vary spatially and temporally, being subject to controls including temperature, altitude, latitude, distance from the coast, and amount of precipitation (Bowen, 2010; Bowen and Good, 2015; Cernusak et al., 2016). More specifically, soil water isotopes are determined by a mixture of individual precipitation events with distinct isotopic signals and are also affected by evaporation, both of which lead to the development of isotopic gradients in soil water with depth (Allison et al., 1983; Liu et al., 2015). Many studies have shown that the δ¹⁸O and δ²H values of root/xylem water can be used to characterize the water sources used by plants (Rothfuss and Javaux, 2017; Wu et al., 2018; Wang et al., 2019; Amin et al., 2020; Zhao et al., 2020; Liu et al., 2021a). These studies rested substantially on the assumption that no isotopic fractionation of δ¹⁸O and δ²H values occurs during water uptake by plant roots (Dawson and Ehleringer, 1991; Ehleringer and Dawson, 1992; Chen et al., 2020), except in saline or xeric environments (Lin and Sternberg, 1993; Ellsworth and Williams, 2007). Some recent studies showed, however, that the occurrence of isotopic fractionation during root water uptake was probably more common than previously thought, especially with respect to δ²H values (Zhao et al., 2016; Wang et al., 2017; Barbeta et al., 2019; Poca et al., 2019; Liu et al., 2021a, 2022).

In addition to plant source water, leaf water is also isotopically enriched through the evaporative process during transpiration. The enrichment of ¹⁸O and ²H by leaf water transpiration can be predicted using the Craig–Gordon model (C–G model). This model was initially proposed to describe the evaporative enrichment of a freely evaporating water body (Craig and Gordon, 1965) and has been modified for plant leaves under steady-state conditions (Dongmann et al., 1974; Farquhar and Cernusak, 2005). However, the C–G model fails to explain the intra-leaf heterogeneity of δ¹⁸O_leaf and δ²H_leaf (Cernusak et al., 2016; Liu et al., 2021b), which is currently described using a two-pool model (Leaney et al., 1985; Song et al., 2015) and/or an advection–diffusion model, as the Péclet effect (Farquhar and Lloyd, 1993; Farquhar and Gan, 2003). Subsequently, more complicated models have been developed to cover non-steady-state conditions (Ogée et al., 2007). These models emphasize a mechanistic understanding of leaf water isotopic fractionation, but the relevant parameters cannot be strictly constrained or precisely monitored, which hinders the use of these models under natural conditions (Plavcová et al., 2018).

This study combined the effects of measured source water isotopes and C–G model-predicted transpiration on δ¹⁸O_leaf and δ²H_leaf values. Our objectives were to deepen the understanding of the controls on the δ¹⁸O_leaf and δ²H_leaf across different seasons. Based upon these objectives, we repeatedly sampled soils, twigs, and leaves in May, July, and September (representing spring, summer, and fall, respectively) from the same 10 plots distributed along an elevation transect. Simultaneously, we obtained the relevant meteorological parameters (e.g., temperature, relative humidity, and precipitation) from sites close to the sampling plots along the transect and used these to predict the δ¹⁸O_leaf and δ²H_leaf values. The combined analysis of concurrent measurements of δ¹⁸O and δ²H values in soil water, twig water, and leaf water with the predicted δ¹⁸O and δ²H values of leaf water from the C–G model associated with the surrounding meteorological parameters will help to identify the factors that control δ¹⁸O_leaf and δ²H_leaf values. Furthermore, we performed an isotope-based line analysis of the dual δ¹⁸O and δ²H values of leaf water, associated with altitude and seasonality. This study will improve our understanding of the environmental signals preserved within the δ¹⁸O and δ²H values extracted from plant organic biomarkers associated with leaf water.

2 Materials and methods

2.1 Study area

The Qinling Mountains form the dividing line between northern and southern China and mark the boundary between the watersheds of the Yellow and Yangtze rivers. Mt. Taibai (Fig. 1; 33.96^∘ N, 107.77^∘ E) rises to 3767 m a.s.l. (above sea level) and is the peak in the Qinling Mountains; it has a warm temperate ecosystem characterized by a rich diversity of flora and fauna. The mean annual temperature at the bottom of Mt. Taibai is 12.9 ^∘C, and the mean annual precipitation is 609.5 mm (Zhang and Liu, 2010). The climate, soil, and vegetation vary significantly along our slope transect, exhibiting a remarkable vertical geo-ecological zonation (Fig. 1). The area contains a variety of climate zones: warm temperate (<1300 m a.s.l.), temperate (1300–2600 m a.s.l.), cool temperate (2600–3350 m a.s.l.), and alpine (>3350 m a.s.l.). The soil types vary from yellow loess soil at low elevations, spectacular rocky outcrops at middle elevations, and glacial remnants at high elevations. Vegetation along the transect is mainly coniferous and broadleaf forests, as well as alpine and subalpine vegetation (Fig. 1; Liu, 2021). The dominant species range from Quercus variabilis, Q. aliena, Betula albosinensis, B. utilis, Abies fargessi, and Larix chinensis forests to Rhododendron clementinae and R. concinnum alpines (Fig. S1 in the Supplement).

Figure 1 Sample sites (red dots) and weather stations (open triangles) that are distributed along vertical vegetation zones across the Mt. Taibai transect on the Chinese Loess Plateau (a). The meteorological parameters (precipitation, temperature, and relative humidity) vary with stations along the elevation transect (b). Mean annual (MAP, MAT, MARH) and monthly (MMP, MMT, MMRH) precipitation, temperature, and relative humidity. The subscripts refer to the month. The vertical vegetation distribution was adopted from Liu (2021).

2.2 Sampling strategy

Plants and soils were sampled in May, July, and September 2020, and samples were collected from 10 plots (3×3 m) covering all the vegetation zones along the northern slope of Mt. Taibai, extending from 608 to 3533 m a.s.l. (Fig. 1). Among the plots, six sites (i.e., sites 2, 3, 4, 5, 8, and 10; Fig. 1) were selected as being the closest to the weather stations along the elevation transect, and they were used in order to obtain the in situ meteorological data for analysis. For the plants, one or two dominant deciduous and coniferous trees were chosen in each plot across the vegetation zone (Fig. S1). Several large leaves and suberized twigs were collected for each species. Between 3 to 10 large leaves were chosen for sampling, and a small number was collected in broadleaf forests and a large number in coniferous forests, depending on leaf size. The leaf samples were conducted in the context of the intact leaves because of the likely isotopic gradients within a leaf (Helliker and Ehleringer, 2000; Liu et al., 2016). Our sampling period was between 12:00 and 15:00 CST, because maximum diurnal enrichment of the leaf water isotopic composition occurs during this part of the day (Romero and Feakins, 2011; Liu et al., 2021). The twigs were collected simultaneously by cutting suberized twigs, and all of the twigs were cut into samples that were 3–4 cm long. The leaf and twig samples were immediately placed into glass vials with screw caps and sealed with polyethylene parafilm. For the soils, three surface soil samples (less than 10 cm deep) were collected from around the sampled plants using a small metal scoop at each plot. All sampling plots were located on slopes far from rivers and surface water bodies, which ensured that the soil water in each plot was derived exclusively from precipitation. Although the surface soil layers were collected only as the representatives of soil water in this study, these samples could provide a relatively good source of water for the plants, as supported by a prior study conducted along the same elevation transect (Zhang and Liu, 2010). The soil samples were tightly sealed in a polyethylene zipper bag on-site. All plant and soil samples were stored in a cool box (∼4 ^∘C) in the field and immediately transported to the laboratory. The altitude of each plot was determined using a handheld GPS unit with an error of ±5 m.

2.3 Isotope analysis

The water in the plant and soil samples was extracted using an automatic cryogenic vacuum extraction system (LI-2100 Pro, LICA United Technology Limited, Beijing, China). The auto-extraction process was set for 3 h, and the extraction rate of water from the samples was more than 98 %. The isotopic composition of the soil water was measured using a Picarro L2130-i isotope water analyzer (Sunnyvale, CA, USA) at the State Key Laboratory of Loess and Quaternary Geology, Institute of Earth Environment, Chinese Academy of Sciences. The analytical accuracies were ±0.1 ‰ for δ¹⁸O and ±1 ‰ for δ²H. An isotope ratio mass spectrometer was coupled to a high-temperature conversion elemental analyzer (HT2000 EA-IRMS, Delta V Advantage; Thermo Fisher Scientific, Inc. USA) to take isotopic measurements of twig and leaf water at the Huake Precision Stable Isotope Laboratory on the campus of Tsinghua Shenzhen International Graduate School. The measurement precisions were ±0.2 ‰ and ±1 ‰ for δ¹⁸O and δ²H, respectively. The isotopic composition of δ¹⁸O and δ²H is expressed as an isotopic ratio:

$$\begin{array}{}\text{(1)}& {\mathit{\delta }}_{\mathrm{sample}}\left(\mathrm{‰}\right)=\left({\displaystyle \frac{R_{\mathrm{sample}}-R_{\mathrm{standard}}}{R_{\mathrm{standard}}}}\right)\times \mathrm{1000},\end{array}$$

where δ_sample represents δ¹⁸O or δ²H, and R_sample and R_standard indicate the ratio of ${}^{\mathrm{18}}\mathrm{O}/{}^{\mathrm{16}}\mathrm{O}$ or ${}^{\mathrm{2}}\mathrm{H}/{}^{\mathrm{1}}H$ of the sample and standard, respectively. The δ¹⁸O and δ²H values are reported relative to the Vienna Standard Mean Ocean Water (VSMOW). In addition, the mean monthly δ¹⁸O and δ²H values of precipitation were determined using the Online Isotopes in Precipitation Calculator (Bowen and Revenaugh, 2003).

2.4 Modeling isotopes of leaf water

The C–G equation can be approximated as follows (Cernusak et al., 2022):

$$\begin{array}{}\text{(2)}& {\mathit{\delta }}_{\mathrm{e}}={\mathit{\delta }}_{\mathrm{s}}+{\mathit{\epsilon }}^{+}+{\mathit{\epsilon }}_{\mathrm{k}}+\left({\mathit{\delta }}_{\mathrm{v}}-{\mathit{\delta }}_{\mathrm{s}}-{\mathit{\epsilon }}_{\mathrm{k}}\right)\times {\displaystyle \frac{e_{\mathrm{a}}}{e_{\mathrm{i}}}},\end{array}$$

where δ_e is the predicted δ¹⁸O and δ²H values at the evaporative sites within leaves, δ_s is the δ¹⁸O and δ²H values of source water (equivalent to twig water in our study), ε⁺ is the equilibrium fractionation between liquid water and vapor, and ε_k is the kinetic fractionation during the diffusion of vapor through the stomata and the boundary layer.

In our analysis, we calculated Δ_v (the enrichment of atmospheric vapor relative to source water) as ${\mathrm{\Delta }}_{\mathrm{v}}=({\mathit{\delta }}_{\mathrm{v}}-{\mathit{\delta }}_{\mathrm{s}})/(\mathrm{1}+{\mathit{\delta }}_{\mathrm{s}})$, and the values of Δ_v are often close to $-{\mathit{\epsilon }}^{+}$ at the isotopic steady state (Barbour, 2007; Cernusak et al., 2016); therefore, we can calculate δ_v as ${\mathit{\delta }}_{\mathrm{v}}=-{\mathit{\epsilon }}^{+}+(\mathrm{1}-{\mathit{\epsilon }}^{+}){\mathit{\delta }}_{\mathrm{s}}$. In addition, $\frac{e_{\mathrm{a}}}{e_{\mathrm{i}}}$ is the ratio of the water vapor pressure fraction in the air relative to that in the intercellular spaces and is equal to the relative humidity (RH) in the air at steady state (Cernusak et al., 2022). Thus, Eq. (2) can be derived as

$$\begin{array}{}\text{(3)}& {\mathit{\delta }}_{\mathrm{e}}=(\mathrm{1}-h)\left({\mathit{\epsilon }}^{+}+{\mathit{\epsilon }}_{\mathrm{k}}\right)+\left(\mathrm{1}-{\mathit{\epsilon }}^{+}h\right){\mathit{\delta }}_{\mathrm{s}},\end{array}$$

where δ_s represents the isotopic values of twig water, and h is the mean annual or monthly RH (MARH or MMRH) in this study. The equilibrium fractionation (ε⁺) varies as a function of temperature (Bottinga and Craig, 1969) and can be equated to δ¹⁸O and δ²H, as follows (Majoube, 1971):

$$\begin{array}{}\text{(4)}& {\displaystyle }\begin{array}{rl}{\mathit{\epsilon }}_{\mathrm{O}}^{+}\left(\mathrm{‰}\right)& =\left[\exp \left({\displaystyle \frac{\mathrm{1.137}}{(\mathrm{273}+T{)}^{\mathrm{2}}}}\times {\mathrm{10}}^{\mathrm{3}}-{\displaystyle \frac{\mathrm{0.4156}}{\mathrm{273}+T}}\right.\right.\\ & \left.\left.-\mathrm{2.0667}\times {\mathrm{10}}^{-\mathrm{3}}\right)-\mathrm{1}\right]\times \mathrm{1000},\end{array}\text{(5)}& {\displaystyle }\begin{array}{rl}{\mathit{\epsilon }}_{\mathrm{H}}^{+}\left(\mathrm{‰}\right)& =\left[\exp \left({\displaystyle \frac{\mathrm{24.844}}{(\mathrm{273}+T{)}^{\mathrm{2}}}}\times {\mathrm{10}}^{\mathrm{3}}-{\displaystyle \frac{\mathrm{76.248}}{\mathrm{273}+T}}\right.\right.\\ & \left.\left.+\mathrm{52.612}\times {\mathrm{10}}^{-\mathrm{3}}\right)-\mathrm{1}\right]\times \mathrm{1000}.\end{array}\end{array}$$

The kinetic fractionation (ε_k) can be calculated for δ¹⁸O and δ²H as (Farquhar et al., 2007; Cernusak et al., 2016)

$$\begin{array}{}\text{(6)}& {\displaystyle }& {\displaystyle }{\mathit{\epsilon }}_{\mathrm{k}}^{\mathrm{O}}\left(\mathrm{‰}\right)={\displaystyle \frac{\mathrm{28}{r}_{\mathrm{s}}+\mathrm{19}{r}_{\mathrm{b}}}{r_{\mathrm{s}}+r_{\mathrm{b}}}},\text{(7)}& {\displaystyle }& {\displaystyle }{\mathit{\epsilon }}_{\mathrm{k}}^{\mathrm{H}}\left(\mathrm{‰}\right)={\displaystyle \frac{\mathrm{25}{r}_{\mathrm{s}}+\mathrm{17}{r}_{\mathrm{b}}}{r_{\mathrm{s}}+r_{\mathrm{b}}}},\end{array}$$

where r_s and r_b are the resistances of the stomatal and boundary layers, respectively (i.e., the inverse of the conductance of the stomatal and boundary layers). Previous studies have found stomatal and boundary layer conductance values of 0.49 and 2.85 mol m⁻² s⁻¹, respectively (Cernusak et al., 2016; Munksgaard et al., 2017), resulting in ${\mathit{\epsilon }}_{\mathrm{k}}^{\mathrm{O}}$ and ${\mathit{\epsilon }}_{\mathrm{k}}^{\mathrm{H}}$ values of 26.7 and 23.8, respectively.

2.5 Statistical analysis

Statistical analysis (i.e., the mean, maximum, and minimum values, as well as the standard deviation) of the isotopes extracted from the precipitation, soil, twig, and leaf samples was performed to define the range and distribution of the δ¹⁸O and δ²H values across the seasons. The Pearson correlation method was used to assess the correlations between the δ¹⁸O and δ²H values among the different water types (i.e., precipitation, soil water, twig water, and leaf water). Hierarchical cluster analysis was used to show the relationships among δ¹⁸O_leaf and δ²H_leaf values and potential source water isotopes (δ¹⁸O and δ²H values in precipitation, soil water, twig water, and leaf water), and meteorological parameters such as mean annual and monthly precipitation (MAP and MMP), mean annual and monthly temperature (MAT and MMT), and mean annual and monthly relative humidity (MARH and MMRH). A one-way analysis of variance (ANOVA) combined with a post hoc Tukey's least significant difference (LSD) test were performed to identify the significant differences in the isotopic compositions of precipitation, soil, twig, and leaf waters across the months. Comparisons of the relationships of δ¹⁸O and δ²H in the soil and leaf water were performed using covariance analysis (ANCOVA) to compare slopes across months. The structural equation model (SEM) was used to explain the respective effects of source waters (i.e., twig water, soil water, and precipitation) and meteorological parameters (i.e., temperature, precipitation, and RH) on δ¹⁸O_leaf and δ²H_leaf values. The validated SEMs generated a good model fit, as indicated by a non-significant χ² test (p>0.05), a high comparative fit index (CFI > 0.95), and a low root mean square error of approximation (RMSEA < 0.05). A special SEM was constructed based on the Mantel R values in AMOS (version 24.0.0). Moreover, we used the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Rolph, 2003) to calculate air mass back trajectory for the central site (34.13^∘ N, 107.83^∘ E, 2270 m a.s.l.) in the study area. These trajectories were initiated four times daily (at 00:00, 06:00, 12:00, and 18:00 CST), and their air parcel was released at 2300 m a.s.l. for May, July, and September 2020 and moved backwards by winds for 120 h (5 d).

3 Results

3.1 Differing response of δ¹⁸O and δ²H values of leaf water

The measured δ¹⁸O and δ²H values of leaf water responded differently to source water isotopes (Fig. 2a) and meteorological parameters (Fig. 2b) across the seasons. The leaf water δ¹⁸O and δ²H values (δ¹⁸O_leaf and δ²H_leaf) were clustered with those of the twig water (δ¹⁸O_twig and δ²H_twig; Fig. 2a) and with MARH, MAT, and MMT (Fig. 2b). The δ²H_leaf values were more closely correlated with isotopes of the potential source waters (e.g., twig water, soil water, and precipitation) than the δ¹⁸O_leaf values in different months (Fig. 2a). In contrast, leaf water δ¹⁸O and δ²H values were correlated with meteorological parameters (Fig. 2b) throughout the study period. These correlations were more significant in summer (July) and autumn (September) than in spring (May).

Figure 2 Heat maps of correlations (r) between leaf water δ¹⁸O and δ²H values and potential source water δ¹⁸O and δ²H values (twig water, soil water, and precipitation δ¹⁸O and δ²H values; a), and meteorological parameters (e.g., MAP, MMP, MAT, MMT, MARH, MMRH). The hierarchical cluster analysis of the isotopes of leaf water and source water (a) and meteorological parameters (b). The subscripts (p, soil, twig, leaf) refer to precipitation, soil water, twig water, and leaf water. ^* Corrected significance at p<0.05, ${}^{**}$ corrected significance at p<0.01, and ${}^{***}$ corrected significance at p<0.001.

3.2 Comparisons of measured and predicted δ¹⁸O and δ²H values of leaf water

The δ¹⁸O_leaf and δ²H_leaf values predicted by the C–G model were compared with the measured δ¹⁸O and δ²H values across all three months (Fig. 3). The C–G model explained 49 % and 70 % of the observed variations in the δ¹⁸O_leaf and δ²H_leaf values, respectively (Fig. 3a and c). The slopes of the relationships for both δ¹⁸O and δ²H values of leaf water were less than one, which suggests that part of the bulk leaf water is derived from unenriched vein water. However, there were no significant differences in δ¹⁸O_leaf (p=0.54; Fig. 3b) and δ²H_leaf values (p=0.93; Fig. 3d) between the C–G model predicted values and the measured values.

Figure 3 Measured leaf water isotopic composition for δ¹⁸O (a) and δ²H (c) values against values predicted by the C–G model. Boxplots show no significant differences for δ¹⁸O (b) and δ²H (d) values between measured and predicted leaf water. The dotted lines show one-to-one lines.

3.3 Variations of δ¹⁸O and δ²H values of different waters with seasons and altitude

There was a significant correlation between δ¹⁸O_leaf and δ²H_leaf values (R²=0.81, p<0.01; Fig. 4), with significant clusters of δ¹⁸O_leaf and δ²H_leaf values across the months, and values were higher in May, intermediate in July, and lower in September (Fig. 4). Within each month, the δ¹⁸O_leaf and δ²H_leaf values were depleted in ²H and ¹⁸O at higher altitudes relative to lower altitudes. Likewise, the potential types of source water (i.e., twig water, soil water, and precipitation) exhibited consistent variations across the months, showing values that were relatively higher in May, intermediate in July, and lower in September (Fig. S1). The correlations between δ¹⁸O and δ²H values among the source waters were also significant (Fig. S2). Nevertheless, the slopes and coefficients of determination (R²) between the δ¹⁸O and δ²H values showed a decrease for precipitation, soil water, twig water, and leaf water from the three sampling months, except for soil water in May (Fig. S2). In addition, the ANCOVA showed no significant differences for the regression lines for precipitation (df=0.47, F=2.49, $p=\mathrm{0.11}>\mathrm{0.05}$), twig water (df=53.2, F=0.42, $p=\mathrm{0.66}>\mathrm{0.05}$), and leaf water (df=437.3, F=2.78, $p=\mathrm{0.08}>\mathrm{0.05}$) across the study months but a significant difference for soil water across the months (df=308.8, F=10.9, p<0.05).

Figure 4 Correlation of leaf water δ¹⁸O and δ²H values across months and altitude. Leaf water δ¹⁸O and δ²H values were higher in May, intermediate in July, and lower in September, and while within each month, those isotopic values were relatively lower at high altitudes and higher in lower altitudes.

4 Discussion

4.1 δ¹⁸O and δ²H values of leaf water

A recent global meta-analysis indicated that δ¹⁸O_leaf and δ²H_leaf values reflect environmental drivers differently and showed that δ²H_leaf values more strongly reflect xylem water and atmospheric vapor δ²H values, whereas δ¹⁸O_leaf values more strongly reflect air relative humidity (Cernusak et al., 2022). Seasonal and localized observations along an elevation transect on the Chinese Loess Plateau supported these different responses of δ¹⁸O_leaf and δ²H_leaf to the isotopic composition of the source water and meteorological conditions (Fig. 2). This is likely due to the fact that variation in precipitation isotopic values compared with that in leaf water evaporative enrichment is larger for δ²H_leaf than δ¹⁸O_leaf (Cernusak et al., 2022). In addition, we found stronger correlations between δ²H_leaf and isotope values of the source water (twig water, soil water, and precipitation) than between δ¹⁸O_leaf values and the source water isotope values (Fig. 2a). This is consistent with the global meta-analysis results (Cernusak et al., 2022). However, our localized study did not show a significantly different response of δ¹⁸O_leaf and δ²H_leaf values to meteorological parameters, which responded at almost equivalent magnitudes (Fig. 2b). These observations suggest that plant organic isotopic proxies such as leaf wax (Sachse et al., 2012; Liu et al., 2016) and cellulose (Barbour, 2007; Lehman et al., 2017), which originate from δ¹⁸O_leaf and δ²H_leaf values, can provide comparative information that indicates climatic signals (e.g., temperature, RH, and precipitation) in natural archives. These results argued with the recent global meta-analysis that δ¹⁸O_leaf and δ²H_leaf values reflect climatic parameters (i.e., RH and temperature) differently (Cernusak et al., 2022). The stronger correlations for δ²H_leaf values than δ¹⁸O_leaf values with isotopic values of the source water were likely, because the δ²H_leaf values are ultimately determined only by precipitation δ²H (Sachse et al., 2012; Liu et al., 2016), whereas the δ¹⁸O_leaf values are affected by a mixture of precipitation δ¹⁸O and atmospheric factors (O₂ and CO₂) (Barbour, 2007; Cernusak et al., 2016). However, the comparative responses of both δ²H_leaf and δ¹⁸O_leaf values to climatic parameters were probably due to the same conditions surrounding the leaf.

The results of the cluster analysis showed that the isotope values of leaf water (δ¹⁸O_leaf and δ²H_leaf) and twig water (δ¹⁸O_twig and δ²H_twig) were clustered into one group, but those of soil water (δ¹⁸O_soil and δ²H_soil) and precipitation (δ¹⁸O_p and δ²H_p) were clustered into another (Fig. 2a). This indicates that the direct source water of δ¹⁸O_leaf and δ²H_leaf should be δ¹⁸O_twig and δ²H_twig, providing the source water isotope basis for the C–G model. In the C–G model (see Eq. 2), besides the source water isotopes, the equilibrium fractionation factor (ε⁺) and atmospheric vapor enrichment (Δ_v) depend on the temperature at the isotopic steady state. Thus, the δ¹⁸O_leaf and δ²H_leaf values were predicted to be associated primarily with temperature, RH, and source water, which is consistent with the results from the cluster analysis that the δ¹⁸O_leaf and δ²H_leaf values were clustered with temperature (MAT and MMT) and RH (MARH; Fig. 2b). Based on the C–G model, we plotted the measured and predicted δ¹⁸O_leaf and δ²H_leaf values (Fig. 3a and c) and observed no significant differences between them (Fig. 3b and d). This is because our three-repeated samplings occur during the day when leaf water is generally near an isotopic steady state when chloroplasts are mostly located near the evaporative sites (Cernusak et al., 2016). The non-steady-state effects on leaf water isotopes were expected at night because of low stomatal conductance (Cernusak et al., 2005, 2016; Cuntz et al., 2007). Although the slopes of the predicted and measured δ¹⁸O_leaf and δ²H_leaf values were less than one, the C–G model still provides a reasonable framework for guiding the analysis of the different controls on δ¹⁸O_leaf and δ²H_leaf values.

4.2 Dual δ¹⁸O and δ²H plots of leaf water

There was a significant linear correlation between the δ¹⁸O_leaf and δ²H_leaf values, with remarkable clusters associated with the three months studied (Fig. 4). As is well known, the local meteoric water line (LMWL), generated by precipitation δ¹⁸O and δ²H values at the local scale, serves as an important reference line for intercomparisons of different waters. Furthermore, the regression lines of the δ¹⁸O and δ²H values from soil water, twig water, and leaf water (Fig. S2) suggest that the leaf water isotopes could well inherit isotopic signals of source waters that originate from twig water, soil water, and ultimately precipitation. The slopes and intercepts of the δ¹⁸O and δ²H values decreased significantly from precipitation, soil water, twig water, and leaf water for each month, except for soil water in May (Fig. S2). Such patterns have been observed in many previous calibration studies (Brooks et al., 2010; Evaristo et al., 2015; Sprenger et al., 2016, 2017; Wang et al., 2017; Benettin et al., 2018; Barbeta et al., 2019; Penna and Van Meerveld, 2019; Liu et al., 2021a, 2022). The slopes of the LMWLs were lower in July (6.79) than in May (7.04) and September (6.85) but were not significantly different (ANCOVA test: df=0.47, F=2.49, $p=\mathrm{0.11}>\mathrm{0.05}$). This suggests that the local water vapor from precipitation was derived from the same source across the seasons but was subject to different intensities of evaporation as the temperature changed throughout the seasons (Li et al., 2019; Wu et al., 2019, 2021). The slopes of the δ¹⁸O and δ²H values from the soil, twig, and leaf waters were much smaller than the LMWLs across the study months due to the secondary evaporation in the other water types.

In the dual isotope plot of leaf water, there were well-defined clusters of δ¹⁸O_leaf and δ²H_leaf values across the three months: ¹⁸O and ²H were depleted in September, there were intermediate values in July, and ¹⁸O and ²H were enriched in May (Fig. 4). When focusing on each month, relatively higher isotopic values occurred at low elevations, but lower isotopic values were present at high elevations despite there being no, or only weak, correlations between the δ¹⁸O_leaf and δ²H_leaf values and altitude (Fig. S3). The correlations between the δ¹⁸O_leaf and δ²H_leaf values and altitude, and between the δ¹⁸O_twig and δ²H_twig values and altitude, were not significant across the three months; however, the δ¹⁸O_p and δ²H_p, and also the δ¹⁸O_soil and δ²H_soil values, were significantly correlated with altitude (Fig. S3), indicating that besides source water (precipitation and soil water), other factors associated with plants also affect δ¹⁸O_leaf and δ²H_leaf values.

The dual isotope plot of δ¹⁸O_leaf and δ²H_leaf values shows a significant isotope line, $y=\mathrm{4.52}x-\mathrm{50.7}$ (R²=0.81, p<0.01; Fig. 4), but shallower slopes (3.53, 1.86, and 2.81 in May, July, and September, respectively) of δ¹⁸O_leaf and δ²H_leaf values were observed across the seasons (Fig. S2). Such a correlation was supported by a recent study that conducted consecutive measurements of δ¹⁸O and δ²H values in xylem and leaf water in Switzerland and indicated that leaf water provided great potential to determine the source water of plants (Benettin et al., 2021). Our study showed remarkable clusters in the measured (Fig. 4) and the C–G model-predicted (Fig. 3) δ¹⁸O_leaf and δ²H_leaf values across the months and the consistencies of respective δ¹⁸O_leaf and δ²H_leaf values with potential source water isotopes across months (Fig. S1). These findings of temporally consistent dynamics among the water types (i.e., precipitation, soil water, twig/stem water, and leaf water) have been observed in a number of previous studies (Phillips and Ehleringer, 1995; Cernusak et al., 2005; Sprenger et al., 2016; Berry et al., 2017; Liu et al., 2021a).

Figure 5 Variation of monthly mean precipitation δ¹⁸O (a) and δ²H (b) values at Xi'an station from the Global Network of Isotopes in Precipitation (GNIP) and cluster mean of moisture transport routes using the HYSPLIT model in May (c), July (d) and September (e), 2020. The background in (c)–(e) is the average precipitation (mm d⁻¹) and 850 hPa wind vectors (arrows, m s⁻¹) in May (c), July (d), and September (e) in AD 1979–2016 based on the database of the Global Precipitation Climatology Center (GPCC) (Becker et al., 2011) and the Modern-Era Retrospective analysis for Research and Applications (Rienecker et al., 2011).

The isotopic inheritance from precipitation to leaf water indicates that seasonal variations of δ¹⁸O_p and δ²H_p values are the first-order control on the temporal patterns observed in leaf water. The seasonal dynamics of the δ¹⁸O_p and δ²H_p values reflect the combined effects of factors such as temperature, altitude, and precipitation amount, which are associated with orographic conditions, as well as sub-cloud evaporation, moisture recycling, and differences in the vapor source (Dansgaard, 1964; McGuire and McDonnell, 2007; Li et al., 2016; Penna and Van Meerveld, 2019; Wu et al., 2019). In this study, we used the HYSPLIT model to demonstrate the ultimate cause of the seasonal variations of δ¹⁸O_leaf and δ²H_leaf values, that is, the monthly dynamics of the δ¹⁸O_p and δ²H_p values. The monthly variations of the δ¹⁸O_p and δ²H_p values from the Global Network for Isotopes in Precipitation (GNIP, http://www.iaea.org/, last access: 18 April 2021) at Xi'an station (AD 1985–1992), which is ∼100 km from our study transect, were enriched in ¹⁸O and ²H in May relative to July and September (Fig. 5a and b). The cluster mean of the moisture transport routes from HYSPLIT (Draxler and Rolph, 2003) and the climatological 850 hPa wind vectors showed that the primary moisture sources were from western China and central Asia in May, the China–India Peninsula and the Bay of Bengal, and local moisture recycling and convection (Fig. 5c–e). The seasonal variations in δ¹⁸O_p and δ²H_p values are consistently related to the onset, advancement, and retreat of the Asian summer monsoon and associated changes in the large-scale monsoon circulation (e.g., Zhang et al., 2020, 2021). As the summer monsoon starts in mid-May, the rainfall season starts in southern China; however, our study area is controlled mainly by moisture from the westerlies (Chiang et al., 2015) with relatively higher vapor, δ¹⁸O_p, and δ²H_p values (Fig. 5a–c). In July, the summer monsoon reaches its strongest phase, and the rainfall belt shifts to central and northern China, where the southerly wind brings plenty of moisture from the China–India Peninsula and the Bay of Bengal with lower vapor, δ¹⁸O_p, and δ²H_p values (Fig. 5a, b, and d). When the summer monsoon withdraws in September, the study area is controlled mainly by local moisture recycling and convection (Fig. 5e). Soil water, stored after the June–August monsoon rainfall with its lower δ¹⁸O and δ²H values, results in even lower δ¹⁸O_p and δ²H_p values in September than in July (Fig. S1), causing significantly lower δ¹⁸O and δ²H values of leaf water (Fig. 4).

4.3 Framework of controls for δ¹⁸O and δ²H values of leaf water

To delineate the mechanisms that control the δ¹⁸O_leaf and δ²H_leaf values, we used the SEMs to quantify the complex interactions among δ¹⁸O_leaf or δ²H_leaf values, source waters, and meteorological parameters (Fig. 6). The coefficients of determination (R²) were 0.48 and 0.71 for the δ¹⁸O_leaf and δ²H_leaf values, respectively, indicating that the models explained more variance for δ²H_leaf values than δ¹⁸O_leaf values (Fig. 6). The SEMs showed that potential source waters (i.e., twig water, soil water, and precipitation) had stronger effects on δ²H_leaf relative to δ¹⁸O_leaf values, while the meteorological parameters showed weak effects on both δ¹⁸O_leaf and δ²H_leaf values (a little larger for δ²H_leaf than δ¹⁸O_leaf values). This is consistent with our above correlation analysis (Fig. 2). Surprisingly, the MMT had significant effects on δ¹⁸O_p and δ²H_p values, suggesting that temperature plays a key role in determining δ¹⁸O_p and δ²H_p values, but this finding is not discussed further here. Collectively, the SEMs also showed that source water exerts the first-order control but affects δ¹⁸O_leaf and δ²H_leaf differently; the meteorological parameters had a weak control on δ¹⁸O_leaf and δ²H_leaf, with a more substantial effect on δ²H_leaf than δ¹⁸O_leaf values.

Figure 6 Structural equation model (SEM) of leaf water δ¹⁸O (a) and δ²H (b) values. The structural equation models considered all plausible pathways. Solid lines indicate significant positive (red) or negative (blue) effects, and dashed lines indicate non-significant effects. Grey lines indicate correlations between two variables. Numbers on the arrow indicate significant standardized path coefficients, proportional to the arrow width. The coefficients of determination (R²) represent the proportion of variance explained by the model.

A schematic representation of the controls on δ¹⁸O_leaf and δ²H_leaf values (respective and dual) is shown in Fig. 7. It involves multiple processes associated with the hydroclimatic and biochemical factors that affect δ¹⁸O_leaf and δ²H_leaf values. The meteorological parameters (temperature, RH, and precipitation) exerted distinct effects on the δ¹⁸O and δ²H values of the source water and, thus, on the δ¹⁸O_leaf and δ²H_leaf values, as demonstrated above by the SEM. Significant isotopic fractionation occurred mainly at two key locations across the vertical soil profiles and leaf architectures from precipitation to leaf water. First, an isotopic gradient across the vertical soil profile appeared because of evaporation from the surface soil layers (Ehleringer and Dawson, 1992; Goldsmith et al., 2012; Evaristo et al., 2015). This evaporative isotopic fractionation causes a linearly isotopic trajectory down the soil profile (Goldsmith et al., 2012; Rothfuss and Javaux, 2017; Wu et al., 2018; Wang et al., 2019; Amin et al., 2020; Zhao et al., 2020; Liu et al., 2021a). Second, there were significant isotopic heterogeneities because of transpiration associated with the δ¹⁸O_leaf (Helliker and Ehleringer, 2000; Farquhar and Gan, 2003; Gan et al., 2003; Song et al., 2015) and δ²H_leaf values (Šantrůček et al., 2007; Liu et al., 2016, 2021b) within a leaf, which depends substantially on veinal structures (Liu et al., 2021b). The within-leaf heterogeneity of the δ¹⁸O_leaf and δ²H_leaf values can be explained using the Péclet-modified C–G model (Gan et al., 2003; Farquhar and Gan, 2003; Cernusak et al., 2005, 2016). Collectively, the soil evaporation and leaf transpiration produce isotopic enrichment above source water (precipitation or soil water). Soil evaporation leads to an isotopic gradient across the vertical soil profile, providing water sources for plant root uptake without isotope fractionation during the process (Dawson and Ehleringer, 1991; Ehleringer and Dawson, 1992; Chen et al., 2020). During water transport between roots and leaf petioles, isotopic compositions of xylem water remain unaltered from those in soils (i.e., soil immobile water), until it reaches the leaf, which undergoes water loss (Ehleringer and Dawson, 1992). Within the leaf, transpiration leads to significant isotopic enrichment (Helliker and Ehleinger, 2000; Liu et al., 2015; Cernusak et al., 2016), which is highly dependent on meteorological parameters (e.g., temperature and relative humidity). However, the meteorological parameters varied with altitude and seasonality, yielding an isotopic leaf water line (LWL) in the dual-isotope plot (Fig. 4). The LWL provides an important baseline for leaf-derived organic matter such as cellulose (e.g., Barbour, 2007; Lehman et al., 2017) and leaf wax (Liu et al., 2016, 2021). Overall, the LWL is controlled primarily by altitude and seasonality, as these are the main influencers of the hydroclimatic factors.

Figure 7 Schematics of the respective and dual isotopes of δ¹⁸O and δ²H values from precipitation to leaf water, associated with physical (evaporation at soil profile and transpiration at leaf level) and biochemical processes. The dual isotopes of δ¹⁸O and δ²H values yield an isotopic water line, the slope of which was lower than the LMWL. The intersected angle varied with hydroclimates, associated with altitude and seasonality.

5 Conclusion

Along an elevation transect on the Chinese Loess Plateau, precipitation, soil water, twig water, and leaf water were repeatedly sampled to explore the controls on δ¹⁸O_leaf and δ²H_leaf values associated with meteorological parameters and source water. The effects of meteorological parameters and source water on δ¹⁸O_leaf and δ²H_leaf values were different, and the dual δ¹⁸O_leaf and δ²H_leaf plot generated an isotopic line. We found that δ²H_leaf values were more closely correlated with source water isotopes than δ¹⁸O_leaf values, whereas δ¹⁸O_leaf and δ²H_leaf values were similarly correlated with meteorological parameters along the elevation transect. The observations suggest that plant organic isotopic proxies such as leaf wax and cellulose originating from δ¹⁸O_leaf and δ²H_leaf values can provide comparative climatic information on the Loess Plateau of China. Additionally, the dual-isotope analysis showed that the δ¹⁸O_leaf and δ²H_leaf values were closely correlated because of their similar altitudinal and seasonal responses. The first-order control on δ¹⁸O_leaf and δ²H_leaf values was the source water (i.e., precipitation), and the meteorological parameters had a comparable effect on both δ¹⁸O_leaf and δ²H_leaf values, which varied with altitude and season across the transect on the Loess Plateau. In the future, we will investigate the relationships of intersection angle θ with hydroclimatic and biochemical factors.