Technical note: On uncertainties in plant water isotopic composition following extraction by cryogenic vacuum distillation

Recent studies have challenged the interpretation of plant water isotopes obtained through cryogenic vacuum distillation (CVD) based on observations of a large ²H fractionation. These studies have hypothesized the existence of an H-atom exchange between water and organic tissue during CVD extraction with the magnitude of H exchange related to relative water content of the sample; however, clear evidence is lacking. Here, we systematically tested the uncertainties in the isotopic composition of CVD-extracted water by conducting a series of incubation and rehydration experiments using isotopically depleted water, water at natural isotope abundance, woody materials with exchangeable H, and organic materials without exchangeable H (cellulose triacetate and caffeine). We show that the offsets between hydrogen and oxygen isotope ratios and expected reference values (Δ²H and Δ¹⁸O) have inversely proportional relationships with the absolute amount of water being extracted, i.e. the lower the water amount, the higher the Δ²H and Δ¹⁸O. However, neither Δ²H nor Δ¹⁸O values, were related to sample relative water content. The Δ²H pattern was more pronounced for materials with exchangeable H atoms than with non-exchangeable H atoms. This is caused by the combined effect of H exchange during the incubation of materials in water and isotopic enrichments during evaporation and sublimation that depend on absolute water amount. The H exchange during CVD extraction itself was negligible. Despite these technical issues, we observed that the water amount-dependent patterns were much less pronounced for samples at natural isotope abundance and particularly low when sufficiently high amounts of water were extracted (>600 µL). Our study provides new insights into the mechanisms causing isotope fractionation during CVD extraction of water. The methodological uncertainties can be controlled if large samples of natural isotope abundance are used in ecohydrological studies.

1 Introduction

Hydrogen and oxygen isotope ratios (δ²H and δ¹⁸O) in plant water are powerful tools for tracing the water movement of ecosystems and providing information on the source water used by plants (Goldsmith et al., 2017; Flanagan and Ehleringer, 1991; Brinkmann et al., 2019; Nehemy et al., 2019). The most widely used approach for determining the δ²H and δ¹⁸O of plant water is by first extracting water from plants using cryogenic vacuum distillation (CVD) method (Ehleringer et al., 2000; West et al., 2006; Ingraham and Shadel, 1992; Orlowski et al., 2013). Since no isotopic fractionation generally occurs during plant water uptake (White et al., 1985; Zimmermann et al., 1967; Poca et al., 2019), it is generally believed that the isotope ratios in stem water (root crown water for grasses) derived from the CVD extraction (δ_CVD) are equal to that in the plant source water (δ_source). However, an increasing number of studies have found that δ_CVD, especially ${{\mathit{\delta }}^{\mathrm{2}}\mathrm{H}}_{\text{CVD}}$, does not actually reflect plant source water, as shown by a strong negative δ²H offset (${\mathrm{\Delta }}^{\mathrm{2}}\mathrm{H}={{\mathit{\delta }}^{\mathrm{2}}\mathrm{H}}_{\text{CVD}}-{{\mathit{\delta }}^{\mathrm{2}}\mathrm{H}}_{\text{source}}$) between CVD-extracted water and source water (Lin and Da S. L. Sternberg, 1993; Zhao et al., 2016; Barbeta et al., 2020; Allen and Kirchner, 2022; Newberry et al., 2017). This issue is important because it may invalidate conclusions on plant water uptake patterns and the partitioning of different pools of water in an ecosystem (Allen et al., 2019; Gessler et al., 2022; Goldsmith et al., 2012; Evaristo and Mcdonnell, 2017).

A critical hypothesis for the observed Δ²H is the spatial heterogeneity of hydrogen isotope composition of water in stems (Zhao et al., 2016; Barbeta et al., 2020), i.e. isotopic differences between stem conductive and nonconductive tissues, and isotopic differences from inside to outside along the stem radial direction. This hypothesis was rejected by Chen et al. (2020). The authors rehydrated dry stem samples of saplings of nine plant species in excess of a reference water with a known isotopic composition for 24 h and then CVD-extracted the water that had soaked into the material. They concluded that the observed negative Δ²H was not caused by within-stem isotope heterogeneity, but was more likely related to a hydrogen atom (H) exchange between stem water and the exchangeable H of stem organic tissue that occurs during the CVD extraction. In woody tissue, oxygen-bound H atoms can exchange with those in surrounding liquid water and water vapour. Theoretically, about 30 % to 50 % of the H atoms of cellulose and its precursors are exchangeable (Schuler et al., 2022; Filot et al., 2006). If the CVD extraction disturbs the previously established isotopic equilibration between organic tissue and tissue water because of the gradually isotopically enriched tissue water during extraction, the potential H exchange would affect ecohydrological interpretations of Δ²H (Barbeta et al., 2020; De La Casa et al., 2022). However, Chen et al. (2020) were unable to isolate the H exchange that occurred during the extraction. This is because their experimental design included two possible H exchange steps: one that occurs during rehydration (hereinafter referred to as “H exchange during rehydration”) and a second one that occurs during the extraction itself (hereinafter referred to as “H exchange during extraction”). Only the latter is of interest, because it is the H exchange process that theoretically affects the isotopic composition of CVD-extracted water from actual plant samples. However, empirical evidence for the occurrence of H exchange and its effects on ${{\mathit{\delta }}^{\mathrm{2}}\mathrm{H}}_{\text{CVD}}$ during rehydration and/or extraction is still lacking.

Moreover, the study of Chen et al. (2020) also found a significant positive correlation between Δ²H and sample relative water content (RWC). They posited that the H exchange effect was probably dampened at high RWC. This correlation has also been found for both Δ²H and Δ¹⁸O (i.e. ${{\mathit{\delta }}^{\mathrm{18}}\mathrm{O}}_{\text{CVD}}-{{\mathit{\delta }}^{\mathrm{18}}\mathrm{O}}_{\text{source}}$) in soil samples (Wen et al., 2021). However, using cellulose in wood as an example, much of the O-bound H is engaged in H bonding that links cellulose fibrils together, i.e. the non-freely exchangeable “bridging hydrogen” (Sepall and Mason, 1961; Meier-Augenstein et al., 2014), which requires high temperatures (>100 ^∘C) for accessing (Schuler et al., 2022). As such, the freely exchangeable H in cellulose is theoretically only about 5 % (Sepall and Mason, 1961; Meier-Augenstein et al., 2014). If the magnitude of H exchange effect during extraction is therefore moderate, the observed Δ²H would be more readily attributed to isotopic fractionation and mixing that occur after the water is extracted from the sample. If this were to be true, what really matters would not be the RWC, but rather the absolute amount of water being extracted (hereinafter referred to “absolute water amount (AWA)”). The effects of AWA on extraction uncertainties could influence both Δ²H and Δ¹⁸O through fractionation during CVD extraction. However, no studies have tested the effect of sample water amount on the CVD extraction biases yet, and the CVD extraction uncertainties during liquid–vapour phase changes are not well known.

During the CVD extraction, successive isotope fractionation processes occur at each phase change of the water (liquid–vapour–solid–liquid). When liquid sample water is distilled to water vapour, isotope fractionation is expected if incomplete extraction occurs (Orlowski et al., 2013; West et al., 2006). When the water vapour is trapped by liquid nitrogen and turns into solid ice, a sublimation isotope fractionation is also expected. Studies testing the sublimation of pure water ice at low temperatures into a vacuum have shown that the sublimated percentage of the bulk water aliquot was <10 % within 1 h at −100 ^∘C, and the sublimation rate decreased drastically with the decrease of temperature in the cold trap (Mortimer et al., 2018; Lécuyer et al., 2017). Given the fact that the extracted water is frozen at such a low temperature (−196 ^∘C) during CVD extraction, negligible amounts of water should be released into the vapour phase by sublimation during the extraction. Nevertheless, an observable enrichment of the residual ice is still possible as the water lost to the vapour phase is isotopically depleted (Lécuyer et al., 2017; Mortimer et al., 2018). In addition, isotope fractionation related to evaporation and mixing with water vapour in the laboratory probably occur during the end of the CVD extraction (i.e. mixing of the extracted water with air humidity of the lab air when the water traps are removed from the extraction system). Although this can be partly avoided by flushing the extraction line with N₂ gas, isotope fractionation can occur when the extracted and frozen water thaws in the water collection tube and when it is transferred to storage vials. A better understanding of these individual isotope fractionation processes during CVD extraction would not only aid in identifying the potential sources of methodological biases (Schoppach and Klaus, 2019) and improving the instrumentation, but also contribute to better estimates of plant water sources by mitigating their effects.

To investigate these open questions, we systematically tested for biases induced by CVD extraction of plant water. A series of experiments that involved extracting water from different kinds of materials (with and without exchangeable H) that were incubated or rehydrated using isotopically depleted water and from plant samples at natural isotope abundance were conducted. Specifically, we hypothesized that the Δ²H is influenced by (i) an H-exchange effect, whereas both Δ²H and Δ¹⁸O are influenced by (ii) tissue AWA, and (iii) evaporation and sublimation enrichments. The lower the AWA, the higher the H-exchange effect and evaporation and sublimation fractionation, and the higher the Δ²H and Δ¹⁸O.

2 Materials and methods

2.1 Experiment 1: testing the overall H-exchange effect

For evaluating the overall effect of H exchange during both rehydration and extraction on the Δ²H and Δ¹⁸O values of CVD-extracted water, we conducted an incubation experiment by adding different amounts of strongly depleted reference water (δ²H: −465.9 ‰; δ¹⁸O: −174.2 ‰; hereafter referred to as δ_ref) to the same amount of different dried materials with and without exchangeable H (experiment 1, Fig. 1). The isotopically depleted water was used to make potential isotope effects more evident.

Figure 1 Overview of the experimental designs. Δ, isotopic offset; δ_CVD, isotope composition of CVD-extracted water; δ_ref, isotope composition of the reference water; δ_{ref after rehyd}, isotope composition of the reference water after rehydration; δ_{ref after evap}, isotope composition of the reference water after the evaporation; δ_{CVD_ave}, the average value of isotope composition of the CVD-extracted water; δ_tap, isotope composition of the tap water; RT, room temperature; RH, relative humidity.

The materials with exchangeable H included stem pieces, stem powder, purified stem cellulose powder, and twig pieces. The stem materials were obtained from the xylem of a trunk disc of a mature Larix sibirica grown in Siberia (27 cm diameter, 4 cm thickness). We cut a portion of xylem (a mixture of sap wood and heart wood) from the trunk disc and oven dried it to constant weight before preparing for stem pieces, powder, and cellulose materials. Stem pieces were prepared by cutting the xylem of the disc into 4 mm cubes, while the stem powder was prepared by grinding the same material to a homogeneous powder using a steel-ball mill (MM400, Retsch GmbH, Haan, Germany). The stem cellulose was then extracted from ground stem powder (Schuler et al., 2022). The twig pieces were obtained from five young Larix decidua trees growing in a forest in Birmensdorf, Switzerland (47.36^∘ N, 8.45^∘ E). Two twigs per tree were collected in the morning, outer bark and phloem removed. The twig xylem was oven dried at 60 ^∘C for 24 h then cut into 2 mm pieces.

The materials without exchangeable H were cellulose triacetate and caffeine. The cellulose triacetate (C₄₀H₅₄O₂₇; Sigma-Aldrich, St. Louis, MO, USA, prod. no. 22199) is a white granule. The caffeine material is ≥99 % caffeine anhydrous (C₈H₁₀N₄O₂; Fluka Chemie AG, Buchs, Switzerland, prod. no. 27600) in a white crystalline powder form. All H atoms in cellulose triacetate and caffeine are bound to carbon in a form of methyl group (−CH₃) and thus non-exchangeable with vapour or liquid water. Cellulose triacetate and caffeine are anhydrous chemicals and were stored in a cool/dry place.

For each material and each different water amount (50, 100, 200, 400, 600, 800, and 1200 µL), three replicates were prepared by transferring 200 mg of each material into 12 mL gas-tight glass vials (Exetainer; Labco, Lampeter, UK). The amount of material was chosen because of practical considerations and to generate a range of water/biomass ratios. The materials were oven dried in the vials with caps open at 60 ^∘C for 24 h to remove any remaining moisture in the material. In total, the materials with exchangeable H were oven dried at 60 ^∘C for at least 48 h, and the materials without exchangeable H were oven dried at 60 ^∘C for 24 h because they are dried chemicals and were properly stored. After drying, the vials were closed and the samples were cooled down to room temperature, then we opened the cap to inject the reference water into the vials. Depending on the form of the material, the low water amounts (e.g. 50 and 100 µL) could not fully wet the material, whereas the high water amounts (e.g. >400 µL) oversaturated the material. Then, the vials were sealed and the materials were incubated for 24 h at room temperature. According to Chen et al. (2020), this incubation time is thought to lead to a full exchange of H atoms between materials and reference water. As a control, the experiment was repeated without any material by adding only the range of reference water into the vial. All samples were frozen at −20 ^∘C and then extracted with CVD. Δ was calculated as δ_CVD−δ_ref.

In addition, to evaluate how effective our drying procedure was and how opening the cap for injecting the reference water influenced the dried samples, we conducted a drying test. In the test, 200 mg stem pieces, stem powder, and caffeine with five replicates were transferred into the vials, then caps closed and weighted. We oven dried the samples without caps for 48 h either at 60 or at 105 ^∘C. Every 12 h during the drying, we closed the caps in the oven then transferred the vials to a balance for weighting. After the last weighting, we opened the cap of the vials for 5 s in the lab to simulate the procedure of injecting the reference water. Then the cap closed and samples were weighted again. The percentage of moisture that was removed from the initial sample at different temperatures and hours and the percentage of moisture that was absorbed by the dried sample in the 5 s were calculated (Table S2 in the Supplement).

2.2 Experiment 2: testing the effects of the absolute water amount and relative water content

For separating the H exchange effects on Δ²H and Δ¹⁸O values occurring during rehydration from those occurring during extraction, we extracted water from a range of rehydrated stem segments with different sizes (experiment 2, Fig. 1).

Large stem segments ($\mathrm{10}\mathrm{cm}\times \mathrm{2.5}\mathrm{cm}\times \mathrm{0.3}$ cm) were cut from the L. sibirica disc (see Sect. 2.1) and oven-dried at 60 ^∘C for 24 h before the experiment. We firstly determined the saturated water content of the large stem segments by weighing them before and after they were soaked in excess of deionized water for 24 h. The saturated water content was determined for estimating the AWA of the following smaller stem segment samples. The large stem segments were then oven dried again at 60 ^∘C for 24 h and cut into smaller stem segments with different sizes, aiming to generate a range of samples with a narrow RWC (ca. 40 %), but varying in their AWA from 50–1200 µL. The sizes of the small segments were determined based on the proportional relationship between the size and the saturated water content of the large stem segments. Three replicates were produced for each size. Then, the small stem segments were separately soaked in excess of the isotopic depleted reference water (ca. 25 mL) in sealed glass vials at room temperature for 24 h, following Chen et al. (2020). During the 24 h rehydration, the isotope ratios of the original reference water equilibrated with the exchangeable H in the small stem segments. Thus, by the end of the rehydration, the isotope ratios of water in the small stem segments are assumed equal to the isotope ratio of the reference water after rehydration (δ_{ref after rehyd}) and not to be equal to the original reference water (δ_ref). Subsequently, the small stem segments were taken out of the reference water, the surface water shaken off, and then immediately sealed in vials and frozen at −20 ^∘C to prevent evaporation. A sample of the reference water after the rehydration was taken from each tube for determining δ_{ref after rehyd}. The AWA and the RWC of each small stem segment sample were calculated by comparing the sample weights before and after the extraction. Here, the RWC is defined as $(\text{wet weight}-\text{dry weight})/\text{wet weight}$, its unit is parts per hundred (%). In experiment 2, Δ was calculated as δ_CVD−δ_{ref after rehyd}. This is because it represents the changes in isotope ratios of the water that occurred during the CVD extraction alone.

2.3 Experiment 3: testing sublimation and evaporation effects

For evaluating the potential effects of sublimation and evaporation during CVD extraction on the isotope ratios of the extracted water, different amounts of reference water (50–1200 µL range as above) were added directly into the u-shaped water collection tubes (i.e. in the cold trap) of the CVD extraction system without sample material and then extracted for 2 h following a standard procedure (see Sect. 2.5). The water in the collection tubes was first frozen by the liquid nitrogen cold trap to avoid potential evaporation effects, then the collection tubes were attached to the extraction line for CVD extraction. The experimental design allows us to exclude isotope fractionation related to distillation and condensation occurring when water in the sample vial was relocated to the collection tube (i.e. U-tube), but captures isotope fractionation related to sublimation of the frozen water under the vacuum and evaporation during thawing of water after CVD extraction (experiment 3, Fig. 1). Δ was calculated as δ_CVD−δ_ref.

For isolating the evaporation effect from the sublimation effect, we also added different amounts of reference water (50–1200 µL range as above) into 2 mL glass vials inside a climate chamber (PGR15, Conviron, Manitoba, Canada). The samples were then left to evaporate with lids open at 25 ^∘C and 50 % relative humidity (RH) for 2 h. The samples in the 2 mL glass vials did not require the CVD extraction. The exposed surface area of the water in the 2 mL glass vials was about 64 mm². Three replicates per water amount were performed for both tests (experiment 3, Fig. 1). Notably, CVD extraction was not used for the evaporation test in the climate chamber, thus the Δ was calculated as the difference between the isotope ratios of the reference water after and before the evaporation (i.e. δ_{ref after evap}−δ_ref).

2.4 Experiment 4: extraction tests using samples at isotope natural abundance

In order to test whether the CVD-related isotope fractionation, which was observed with labelled water, can also be seen at natural isotope abundance, we performed extractions with tap water and fresh L. decidua twigs with natural isotope abundance to compare with results derived from tests with isotopically depleted reference water. For the tap water (δ_tap; δ²H: −73.91 ‰, δ¹⁸O: −14.43 ‰), the different amounts of water (i.e. 50, 100, 200, 400, 600, 800, and 1200 µL) were added into the vials and kept frozen at −20 ^∘C until CVD extraction (experiment 4, Fig. 1). The twigs were collected from five young L. decidua trees growing in a forest in Birmensdorf, Switzerland (see Sect. 2.1). The twig xylem was cut to different lengths (1, 2, 3, 4, 5, 6, 7, and 8 cm) and sealed in vials. The AWA and RWC of each twig sample were determined by weighing before and after CVD extraction. Three replicates were performed per water amount and twig length (experiment 4, Fig. 1). For the extraction with tap water, the Δ was calculated as δ_CVD−δ_tap. The isotope ratios of the source water of the L. decidua twigs before extraction were unknown, thus the Δ in this case was calculated as the difference between the δ_CVD and the average value of δ_CVD (i.e. δ_CVD−δ_{CVD_ave}).

2.5 Water extraction and stable isotope analysis

We extracted water using a CVD setup similar to the one described by Orlowski et al. (2013). A schematic overview of this setup is shown in Fig. S1 in the Supplement. During extraction, the pressure inside the system was maintained below 0.05 mbar using a vacuum pump. The sample tubes were additionally blocked by PP fiber filters (Nozzle protection filter, Socorex Isba SA, Ecublens, Switzerland) when water was extracted from powder/pieces of the selected materials to avoid particles being drawn into the U-tubes with the extracted water or the vacuum pump (BS2212, Brook Crompton Ltd, Doncaster, UK). The samples in the sample tubes were heated in 80 ^∘C water, and the extracted water was condensed and trapped in the collection tubes by liquid nitrogen. The extraction was maintained for 2 h to achieve a complete extraction (West et al., 2006). After the extraction, the vacuum inside the system was released by adding dry nitrogen gas until atmospheric pressure conditions were reestablished. Then the collection tubes with frozen water samples were detached from the system and sealed with rubber plugs. The water in the collection tubes was thawed at room temperature. During this process, the evaporated water vapour usually condensed to form very small water droplets on the inside walls of the collection tube. We consolidated and collected as many of these small water droplets as possible and transferred them into glass vials (350 µL or 2 mL, depending on the extracted water amount; Infochroma AG, Goldau, Switzerland) using a pipette. Syringes and 0.45 µm nylon filters were used if the extracted water appeared to be turbid. The samples were stored at −20 ^∘C before and after the extraction.

The δ²H and δ¹⁸O of water samples were measured with a high temperature conversion elemental analyser coupled to a Delta^Plus XP isotope ratio mass spectrometer (TC/EA-IRMS; Finnigan MAT, Bermen, Germany). The δ²H and δ¹⁸O in the materials without exchangeable H (i.e. cellulose triacetate and caffeine) and δ¹⁸O in the bulk organic matter of the materials with exchangeable H were measured with a vario PYRO cube (Elementar Analysensysteme GmbH, Langenselbold, Germany) coupled to a Delta^Plus XP IRMS. Isotope ratios are reported in per mille (‰) relative to Vienna Standard Mean Ocean Water (VSMOW). Calibration versus the international standards was achieved by analysis of a range of certified water of different isotope ratios, resulting in a precision of analyses of 2 ‰ for δ²H and 0.3 ‰ for δ¹⁸O. The non-exchangeable δ²H (δ²H_ne) of the materials with exchangeable H was determined by pre-treating the materials with a high-temperature water vapour equilibration method according to Schuler et al. (2022). The δ²H and δ¹⁸O values of the materials used in the experiments are shown in Table S1 in the Supplement.

2.6 Statistical analyses

To determine if the sample water δ²H and δ¹⁸O values were significantly different before and after the CVD extraction, we performed one-sample t-tests for Δ²H and Δ¹⁸O values to test if they were significantly different from 0. We fit relationships between Δ²H and Δ¹⁸O as a function of AWA using inversely proportional models, and the relationships between Δ²H and Δ¹⁸O as a function of RWC using linear models for the best description of the relationships. The relationships between Δ²H and Δ¹⁸O as a function of RWC were also tested using a linear mixed-effects model with AWA category (AWA<400 or AWA>400 µL) as a random effect with the lmerTest package (Kuznetsova et al., 2017). All statistics were performed using R version 4.0.4 (R Core Team, 2021).

3 Results and discussion

3.1 H exchange occurs between sample exchangeable H and water during rehydration

Experiment 1 revealed an inversely proportional relationship between the AWA and the Δ²H and Δ¹⁸O values of the water derived from material with or without exchangeable H (Fig. 2). In other words, the lower the AWA, the higher the Δ²H and Δ¹⁸O values. For both Δ²H and Δ¹⁸O, the differences among samples at AWA>600 µL were much smaller compared to those at AWA<600 µL. Noticeably, Δ²H of the water derived from materials with exchangeable H showed a more pronounced pattern, which reached about 150 ‰ at 50 µL AWA, compared to those without exchangeable H (Fig. 2a and b). In contrast, the Δ²H pattern of the water derived from materials without exchangeable H was similar to that of the pure reference water, which reached about 30 ‰ at 50 µL AWA (Fig. 2b). For Δ¹⁸O, the patterns were similar among materials with and without exchangeable H, showing an average decrease from 32 ‰ at 50 µL to 1.7 ‰ at 1200 µL (Fig. 2c and d).

The results of experiment 1 suggest that the H exchange during rehydration is mainly responsible for the large Δ²H differences between materials with and without exchangeable H at the low water amounts. This can likely be explained by an isotope mass balance between the exchangeable H in the samples (natural isotope abundance) and H atoms in the reference water (isotopically depleted), with the impact of the exchangeable H of a constant sample size (200 mg) becoming less noticeable as the amount of water increases (from 50 to 1200 µL). Our results thus provide empirical evidence that H exchange occurs during rehydration of material and that this artefact must be considered in rehydration experiments (Chen et al., 2020; Zhao et al., 2022). However, our experiments cannot provide any evidence for the de- and rehydration of plants under natural conditions (i.e. changes in plant relative water content) which can potentially affect isotope composition of plant water (De Deurwaerder et al., 2020; Barbeta et al., 2022). The effect of rehydration under field conditions could be quantified by additional experiments using water with an isotopic composition distinct from that normally expected in soils.

Figure 2 Absolute water amount effects on Δ²H and Δ¹⁸O values (a, b and c, d, respectively) of water derived from materials with and without exchangeable H (a, c and b, d, respectively), of a constant weight (200 mg). Material with exchangeable H differ in their structural composition (pieces vs. powder). “Reference water” reflects a control performed with water but without any material. $\mathrm{\Delta }={\mathit{\delta }}_{\text{CVD}}-{\mathit{\delta }}_{\text{ref}}$ (see Sect. 2.1). The dashed lines represent Δ=0, with grey shading to represent the SD of the reference water isotope analyses (2 ‰ for ${{\mathit{\delta }}^{\mathrm{2}}\mathrm{H}}_{\text{ref}}$; 1.5 ‰ for δ¹⁸O_ref). The curves are inversely proportional fits. Mean values ±1 SD are shown (n=3).

Moreover, the Δ²H of the stem and twig pieces were higher than that of the stem and stem cellulose powder when AWA>600 µL (Fig. 2a), whereas no differences were observed for Δ¹⁸O (Fig. 2c). At a first glance, this is unexpected because woody pieces should be less prone to H exchange due to lower surface area and because the OH groups are locked in the matrix of the woody structure (Sepall and Mason, 1961). A possible explanation for the observed Δ²H discrepancies might be an isotopic difference in the exchangeable H of pieces and powdered samples before the start of the experiment. However, if this was true, the effect should be more visible at lower water amounts rather than higher water amounts, as observed in this study. We therefore can conclude that the sample matrix may also influence H exchange in rehydration experiments, but that the underlying mechanisms remain speculative. Besides, we additionally tested whether the drying procedure (oven drying at 60 ^∘C) in experiment 1 influenced our results. The test showed that all the tested materials were dried to a constant weight after 36 and 12 h at 60 and 105 ^∘C, respectively, despite more moisture being removed at 105 ^∘C (Table S2). Interestingly, after opening the cap for 5 s to simulate injection of reference water, the amount of absorbed lab water vapour relative to the dry weight was about 1 % higher for the samples which were dried at 105 ^∘C than the samples which were dried at 60 ^∘C (Table S2). These results indicate that a complete drying is very difficult in our case, regardless of drying temperature. However, the test provides evidence that only small amounts of remaining moisture remain in the sample after drying and that the reabsorption should not affect the isotopic results of our study, because strongly depleted reference water was used to amplify the isotopic effect.

Unexpectedly, Δ²H values of the pure reference water and of material without exchangeable H, as well as Δ¹⁸O values, all show a dependency on AWA (Fig. 2b–d). This observation indicates that Δ values can be biased even when no exchangeable H in plant material is available, and suggests that other factors than H exchange play an important role in biasing isotopic results of CVD-extracted water.

3.2 Absolute water amount not relative water content causes isotope fractionation during CVD extraction

To better understand the unknown isotopic effects during extraction, we performed experiment 2, where we separated the effects between rehydration and extraction on CVD-extracted water along a similar AWA gradient (50–1600 µL), but in a narrow RWC range (36 %–45 %, Fig. 3). The results confirm the inversely proportional pattern of Δ along an AWA gradient as shown by experiment 1 (Fig. 2). More importantly, the results show that Δ values of water extracted from stem material that has been previously rehydrated were similar (Fig. 3a and c) to that extracted from materials without any exchangeable H (Fig. 2b and d), when the isotope ratio of the reference water after rehydration (δ_{ref after rehyd}) instead of the reference water was considered. This shows that the H-exchange effect during extraction (i.e. all steps after the rehydration procedure) is negligible and cannot be the main reason for observed isotopic offset between expected and measured isotope ratios in water of stem material (Bowers and Williams, 2022).

Figure 3 Effects of absolute water amount and relative water content (a, c and b, d, respectively) on Δ²H and Δ¹⁸O of water derived from the stem segments (a, b and c, d, respectively). The stem segments were rehydrated in excess of the reference water followed by a CVD extraction. $\mathrm{\Delta }={\mathit{\delta }}_{\text{CVD}}-{\mathit{\delta }}_{\text{ref after rehyd}}$ (see Sect. 2.2). The relationship between AWA and Δ was fit with an inversely proportional function; the relationship between RWC and Δ was fit with a linear regression. The dashed lines represent Δ=0. Different colours were used to visually separate data with water amount <400 (red) and >400 µL (blue).

This finding is in contrast to the conclusions made by Chen et al. (2020), stating that stem water CVD extraction error could originate from a dynamic H exchange between sample and water during extraction. The reason for the lack of H-exchange effect during the 2 h CVD extraction itself is probably because the duration for the H exchange was too short to fully exchange. Previous research shows that a large portion of the water in the sample is expected to be extracted within the first 30 min of extraction (Orlowski et al., 2013; West et al., 2006). Thus, there may not be a strong H exchange occurring between the water and sample tissue, especially when a large portion of exchangeable H atoms are not freely accessible at the extraction temperature of 80 ^∘C (Sepall and Mason, 1961). The duration of rehydration might be another factor that potentially affected our results of the rehydration experiments (experiment 1 and 2). According to Chen et al. (2020), a complete equilibration between sample exchangeable H and reference water during rehydration can be achieved within 24 h, given water of a sufficiently large volume. In experiment 2 of our study, the rehydration was also conducted in an excess of reference water for 24 h. The AWA-dependence of Δ²H values of stem segments in experiment 2 (Fig. 3a) was found to be similar to that of the pure reference water and materials without exchangeable H extractions in experiment 1 (Fig. 2b). This likely confirmed that a full equilibrium was achieved during our 24 h rehydration. Nevertheless, the duration, sample size, and amount of water should be considered in future rehydration studies.

We found a weak, positive linear trend in Δ²H with RWC (r²=0.28, p=0.02, Fig. 3b), but not for Δ¹⁸O (p=0.08, Fig. 3d). The trend in Δ²H can be explained by the fact that samples with a smaller AWA (<400 µL) had a ca. 2 % higher RWC compared to those with a higher AWA (>400 µL). The relationships between both Δ²H and Δ¹⁸O as a function of RWC were not statistically significant when AWA was considered as a random factor (AWA with 2 levels: AWA<400 and AWA>400 µL; Fig. 3 and Table S3 in the Supplement). This indicates that the Δ²H and Δ¹⁸O values are dependent on AWA rather than on RWC of the sample. Our result is not consistent with Chen et al. (2020), who found a significant positive correlation of stem RWC with the Δ²H in a rehydration experiment and further recommended determining sample RWC for correcting CVD artefacts. However, our results are well supported by the recent study of Zhao et al. (2022), who CVD-extracted water from rehydrated samples of 12 woody plant species and analysed Δ values. The RWC of their samples ranged from 30 %–60 % and no significant relationships were found between both Δ²H and Δ¹⁸O and RWC. Therefore, the relatively smaller sample RWC range in our study (ca. 10 %) compared to that in Chen et al. (2020), i.e. ca. 20 % for most samples, could not be the reason for the inconsistency of the two studies. Further, given that AWA is typically not reported, we could not test whether the AWA or RWC caused the effect in previous studies. We highlight that the AWA rather than the RWC of a sample should be considered to potentially correct for Δ offsets, and that future studies should report AWA in order to quantify the CVD-induced isotopic biases across laboratories.

3.3 On the processes causing isotope fractionation during CVD extraction

In order to better understand the processes leading to isotope fractionation during CVD, we performed an additional experiment (experiment 3). In this experiment, different amounts of reference water were either injected directly into the collection tubes then subject to CVD extraction, or subject to controlled evaporative conditions in a climate chamber for 2 h. Interestingly, experiment 3 shows that both Δ²H and Δ¹⁸O values of reference water, that was directly transferred to the cold trap (i.e. U-tubes), also followed the inversely proportional pattern as a function of AWA (Fig. 4a). At the lowest AWA, Δ²H reached about 19 ‰, while Δ¹⁸O reached about 9 ‰. For AWA>600 µL, the average Δ²H and Δ¹⁸O were −1.5 ‰ and −2.3 ‰, respectively, and both were significantly different from 0 ($t=-\mathrm{3.13}$ and −3.34, df=8, p<0.01). We suppose that the negative Δ²H and Δ¹⁸O values for the AWA>600 µL were purely caused by the analytical uncertainty, because no incomplete extraction could occur given that the reference water was added directly into the collection tube. The same inversely proportional pattern was observed when water was evaporated under controlled conditions in a climate chamber, whereby Δ²H reached about 46 ‰ and Δ¹⁸O reached about 23 ‰ at the lowest AWA. For AWA>600 µL, average Δ²H was about 3 ‰ and significantly different from 0 (t=12.15, df=8, p<0.01). No significant difference from 0 was found for Δ¹⁸O (t=0.23, df=8, p=0.8). By comparing these results with the one of 50 µL of reference water that was extracted from the sample tube (Fig. 2b), the sublimation and evaporation isotope fractionation account for 59 % and 38 % of the observed Δ²H and Δ¹⁸O, respectively. The unaccounted isotope fractionation may originate from distillation and condensation, when water in the sample tube was relocated to the water collection tube.

Figure 4 Absolute water amount effects on Δ²H (a) and Δ¹⁸O (b) of the reference water that was directly added into the water collection tube before CVD extraction (blue), or evaporated in a climate chamber (red). For the former, $\mathrm{\Delta }={\mathit{\delta }}_{\text{CVD}}-{\mathit{\delta }}_{\text{ref}}$; for the latter, $\mathrm{\Delta }={\mathit{\delta }}_{\text{ref after evap}}-{\mathit{\delta }}_{\text{ref}}$ (see Sect. 2.3). The dashed lines represent Δ=0, with grey shading to represent the SD of the reference water isotope analyses (2 ‰ for ${{\mathit{\delta }}^{\mathrm{2}}\mathrm{H}}_{\text{ref}}$; 1.5 ‰ for δ¹⁸O_ref). The curves are inversely proportional fits. Mean values ±1 SD are shown (n=3).

Two potential reasons can explain this water amount-dependent pattern of Δ. On the one hand, a bi-directional exchange between water droplets and water vapour may lead to an isotopic equilibration between the two sources, similar to observations at the leaf level (Lehmann et al., 2020; Goldsmith et al., 2017). This process is expected to occur when the extracted water drop was thawing inside the water collection tube after extraction. The equilibration effect was presumably large on our extracted water, because of a high isotopic discrepancy between the water vapour of the laboratory (at natural isotope abundance) and the isotopically depleted reference water. The effect is expected to increase as the water pool size decreases, because the smaller the water droplets, the larger the ratio of water vapour volume to water droplets volume in the water collection tube. Therefore, isotope ratios of smaller water amounts were subject to greater influence of the isotopic composition of the laboratory water vapour. On the other hand, the sublimation, evaporation, and/or exchange with surrounding water vapour occurs at the surface of water drops (Stewart, 1975). The smaller the water drop, the larger the ratio of the enriched surface water to the total water drop volume. Thus, the mixing effect of enriched surface water and the rest of the water body would be greater for smaller water drops.

Taken together, our results clearly show that the inversely proportional pattern of Δ along AWA is partly influenced by the isotope fractionation during water sublimation and evaporation during CVD extraction. We therefore conclude that isotope fractionation rather than H-exchange effects with plant material during extraction itself should be considered as factors during CVD extraction of plant water. It should be noted that although our experiments were mainly conducted on samples of Larix, the AWA-dependent patterns of Δ were consistent among extractions with different materials and even pure water. This indicates that the CVD biases we observed on Larix may also be applicable to other taxa. The magnitude of the biases may be slightly different among species (Zhao et al., 2022; Chen et al., 2020; De La Casa et al., 2022), but the AWA-dependency of the biases should show consistency. We also want to highlight that the observed isotope fractionation likely depends on the setup (e.g. system volume) of the CVD extraction system. Therefore, further laboratory comparisons for plant water extraction using different woody and herbaceous plant species are needed for constraining this extraction bias.

3.4 Relevance for isotope fractionation during CVD extraction for samples at natural isotope abundance

The large isotopic discrepancy between the strongly depleted reference water and the plant tissue organic matter allowed us to investigate the processes and mechanisms driving isotope fractionation during CVD extraction. However, plant water at natural isotope abundance is typically more enriched compared to our reference water (δ²H: −465.9 ‰; δ¹⁸O: −174.2 ‰) and therefore the observed isotopic effects of this study are likely less pronounced at natural isotope abundance. This was true, as shown by the results of experiment 4; the inversely proportional pattern was less pronounced or absent with plant samples (average values in δ²H and δ¹⁸O of extracted water: −58.16 ‰ and −7.84 ‰, respectively), or when tap water (δ²H: −73.91 ‰; δ¹⁸O: −14.43 ‰) was used (Fig. 5). For the pure tap water extraction, Δ²H decreased from 3 ‰ at 50 µL to −3.5 ‰ at 1200 µL, while Δ¹⁸O decreased from 2 ‰ at 50 µL to −0.2 ‰ at 1200 µL (Fig. 5a and c).

Figure 5 Absolute water amount and relative water content effects (a, c and b, d, respectively) on Δ²H and Δ¹⁸O (a, b and c, d, respectively) of water at natural isotope abundance. For water of Larix decidua twig xylem, $\mathrm{\Delta }={\mathit{\delta }}_{\text{CVD}}-{\mathit{\delta }}_{\text{CVD\_ave}}$; while for the tap water, $\mathrm{\Delta }={\mathit{\delta }}_{\text{CVD}}-{\mathit{\delta }}_{\text{tap}}$ (see Sect. 2.4). The relationship between water amount and Δ was fit with an inversely proportional function. The dashed lines represent Δ=0. For (a) and (c), mean values ±1 SD are shown (n=3).

The possible explanations for the large difference in the magnitude of the Δ pattern between the reference (Fig. 2b and d) and tap water extractions (Fig. 5a and c) are that the isotope fractionation during CVD extraction is dependent on the initial water isotope ratios and/or the influence from isotope ratios of the surrounding water vapour. Regarding the former, water evaporation tests using water with different isotope ratios (δ²H range: −14.7 ‰ to −57.5 ‰; δ¹⁸O range: 3.4 ‰ to −7.8 ‰) showed that the δ²H and δ¹⁸O values of the initial water have little influence on water evaporation isotope fractionation conducted by Hu et al. (2009). In other words, for water with different initial isotope ratios evaporating in the same conditions, the changes in δ²H or δ¹⁸O would be almost the same. Therefore, the initial water isotope ratios would not be the reason for the significant Δ difference between the reference water and the tap water extraction.

It is therefore possible that the difference in Δ between reference water and tap water extractions is caused by the exchange of extracted water with the water vapour in the laboratory. The isotope ratios of the laboratory water vapour are closer to that of the tap water and twig water, but differ greatly compared to the depleted reference water. Therefore, under these conditions, the initial isotopic signature of the extracted water determines the magnitude of the isotope signature of the extracted water after the exchange with laboratory water vapour and therefore the observed Δ values. This suggests that the observed isotopic variations in our experiments with strongly isotopically depleted reference water might have been less pronounced if water closer to natural isotope abundance would have been used.

For the fresh L. decidua twigs water extraction, the AWA-dependent pattern was evident for Δ¹⁸O, with average values decreasing from approximately 1 ‰ at the lowest water amount to approximately −0.2 ‰ at the highest water amount (Fig. 5c). In contrast, Δ²H ranged between −6.5 ‰ and 6.1 ‰ but without showing a clear inversely proportional pattern (Fig. 5a). While we found no clear explanation for the absence of the AWA dependency for Δ²H in this experiment, we can only speculate that the expected pattern was hidden by processes shaping hydrogen but not oxygen isotopic variations such as H-exchange effects between water vapour and extracted water with similar isotopic compositions. Our results also showed that the Δ values obtained from the extraction of water at natural isotope abundance (Fig. 5a and c) are much smaller compared to those obtained from the extractions of the strongly depleted reference water (Figs. 2 and 3). In addition, we did not find a significant correlation between RWC and Δ²H or Δ¹⁸O for the L. decidua twigs at a RWC range between 45 % and 57 % (Fig. 5b and d). These results suggest again that the Δ²H and Δ¹⁸O of the CVD-extracted water is dependent on AWA rather than on RWC.

4 Conclusions

In conclusion, we provide strong evidence that H exchange with organic H in the sample, despite significantly influencing the δ²H of water during the sample rehydration process, is not the main driver of the broadly observed negative Δ²H value during CVD extraction. Instead, we identified a significant CVD artefact when water is present in small amounts, particularly when δ²H and δ¹⁸O of the water were below natural isotope abundance. This is linked to the increase of sublimation and evaporation enrichments with the decrease of water amount, as well as the mixing between the extracted water and laboratory water vapour, rather than the effect of sample RWC on the H exchange during extraction. From our results, both Δ²H and Δ¹⁸O approached steady values and were close to zero when >600 µL of water was extracted. We therefore recommend extracting more than 600 µL of water, especially for studies using labelled water, to avoid large enrichment biases. However, the minimum amount of extracted water that is necessary to avoid isotopic biases might vary with the CVD setup and therefore laboratory comparison should be conducted. Our results have implications for studies using stable isotopes of water in plant tissue to determine plant water sources (e.g. Evaristo and McDonnell, 2017), trace water through soils with high organic material content (e.g. Koeniger et al., 2016; Sprenger et al., 2016), and reconstruct climate patterns using tree ring tissue (e.g. Loader et al., 2007; Lehmann et al., 2021).