Technical note: Conservative storage of water vapour: a key to practical measurements of water stable isotopes in tree stems and soils

Using water stable isotopes to track plant water uptake or soil water processes has become an invaluable tool in ecohydrology and physiological ecology. Recent studies have shown that laser absorption spectroscopy can measure equilibrated water vapour well enough to support inference of liquid stable isotope composition of plant or soil water, on-site and in real-time. However, current in-situ systems require the presence of an instrument in the field. Here we tested, first in 15 the lab and then in the field, a method for equilibrating, collecting, storing, and finally analysing water vapour for its isotopic composition that does not require an instrument in the field. We developed a vapour storage vial system (VSVS) that relies on in-situ sampling into crimp neck vials with a double-coated cap using a pump and a flow meter powered through a small battery and measuring the samples in a laboratory. All components are inexpensive and commercially available. We tested the system’s ability to store the isotopic composition of its contents by sampling a range of water vapour of known isotopic 20 compositions (from -95 to +1700‰ for δ2H) and measuring the isotopic composition after different storage periods. Samples for the field trial were taken in a boreal forest in northern Sweden. The isotopic composition was maintained to within 0.6 to 4.4‰ for δ2H and 0.6 to 0.8‰ for δ18O for natural-abundance samples. Although 2H-enriched samples showed higher uncertainty, they were sufficient to quantify label amounts. We detected a small change in the isotopic composition of the sample after long storage period, but it was correctable by linear regression models. We observed the same trend for the 25 samples obtained in the field trial for δ18O but observed higher variation in δ2H compared to the lab trial. Our method combines the best of two worlds, sampling many trees in-situ while measuring at high precision in the laboratory. This provides the ecohydrology community a tool that is not only cost-efficient but also easy to use. https://doi.org/10.5194/hess-2022-37 Preprint. Discussion started: 4 February 2022 c © Author(s) 2022. CC BY 4.0 License.


1
Introduction 30 Since the introduction of isotope-ratio infrared spectrometers (IRIS), the analysis of water stable isotope samples has become much more popular in many fields, e.g., in hydrogeologic, watershed, oceanographic or eco(hydro)logical studies (Tweed et al., 2019;Oerter and Bowen, 2017;Oerter et al., 2019). This has led to an increased utility of water stable isotopes also in applications, where the interest of inferring plant water uptake depths/patterns and water movements through the soil matrix has grown tremendously (Eggemeyer et al., 2008;Liu et al., 2010;Beyer et al., 2016;Magh et al., 2020). 35 Until recently, however, samples of matrix-bound water needed to be obtained destructively to extract the water samples.
The state-of-the-art extraction process was, thus far, cryogenic vacuum extraction, where a sample undergoes heating under vacuum, with the bound water evaporating in the process and subsequently being captured in a cryogenic trap (Ingraham and Shadel, 1992;Koeniger et al., 2011;Orlowski et al., 2013Orlowski et al., , 2016. The method was preferred because the assumed completeness of the water extraction was thought to eliminate fractionation. However, it has recently been heavily criticised 40 for introducing biases due to artefacts coming from an exchangeable organic hydrogen pool in the plant biomass (Chen et al., 2020;Allen and Kirchner, 2021).
A recently developed method based on direct vapour equilibration reduces the co-extraction of organic compounds and increases sample throughput (Millar et al., 2018;Wassenaar et al., 2008). One of the biggest advantages of in-situ equilibration techniques is that water from plants and soils can be sampled at high temporal resolution without altering their 45 physiology or physical properties (Kühnhammer et al., 2021). Therefore, in-situ measurements of water stable isotopes have gained popularity and have been proposed a way forward to disentangle isotopic processes in the critical zone or the soilvegetation-atmosphere continuum (Rothfuss and Javaux, 2017).
In-situ measurement systems are based on direct inferences of liquid water isotopic composition from equilibrated water vapour from the soil or the plant (for a detailed review see Beyer et al., (2020). The vapour is collected either using a gas-50 permeable membrane (the utility of which was proven by Herbstritt et al., (2012) buried in the soil (Rothfuss et al., 2013;Volkmann et al., 2016b;Volkmann and Weiler, 2014;Kübert et al., 2020) or in the xylem of woody species (Volkmann et al., 2016a, b;Seeger and Weiler, 2021), or drawing equilibrated water vapour from a borehole in the xylem directly (Marshall et al., 2020;Kühnhammer et al., 2021). Additionally, it is possible to measure the isotopic composition of plant transpiration and evapotranspiration in-situ, using gas exchange chambers in the lab (Simonin et al., 2013;Dubbert et al., 55 2017), as well as in the field (Kübert et al., 2019;Dubbert et al., 2013;Wang et al., 2013).
The biggest advantage of these in-situ systems is their ability to monitor real-time changes in water uptake and subsequent transport in plants and/or in soils and produce immediate data. The biggest disadvantage is the need for an IRIS at the site of measurement, which requires shelter, protection against vandals, and most importantly, access to a continuous power source.
Additionally, the in-situ setup in practice is limited in spatial resolution, as it requires tubing at the length of the distance 60 from the sampling place to the IRIS, which is advisably kept short as increased tubing length increases the possibility of condensation (Beyer et al., 2020). These factors limit the utility of in-situ measurement systems to field sites in vicinity to civil infrastructure, which potentially leads to research sites chosen because of proximity to power rather than suitability as research location, and therefore, location biases (e.g. monitoring wildlife in vicinity to universities (Piccolo et al., 2020), or the location of protected areas worldwide (Joppa and Pfaff, 2009)). Additionally, remote areas tend to lie in regions with less 65 wealth, leading to an underrepresentation of research requiring cost-intensive equipment.
We therefore propose to adapt the above presented in-situ measurement systems to mixed systems, where sample equilibration occurs in-situ but analysis at a central laboratory. This should be useful where in-situ measurements are impossible, due to lack of power supply and safe storage of equipment, or when large numbers of samples or simultaneous observation are a requirement. 70 Here, we introduce an adapted sampling method based on a vacuum pump powered by a 12V battery (derived from the borehole method by Marshall et al. (2020) and a commercially available storage container (adapted from the SWIS System introduced by Havranek et al., (2020), making the presence of an IRIS in the field redundant. We tested our VSVS (Vapour Storage Vial System) using water sources of known isotopic composition in an extensive lab trial and added data from a field trial carried out in a boreal forest in northern Sweden, where we could test the suitability of the proposed method and 75 identify possible limitations. We include a section "preceding work" in the Results section to give the reader a chance to avoid repeating our failures if attempting to improve this methodology.

2
Material and Methods

VSVS lab test 80
We conducted a laboratory test with water of known isotopic composition (i.e. standards). The standards were stored in 50 ml crimp-neck vials (VWR1548-2092, VWR International AB, Stockholm, Sweden). The vials were dried in the oven at 65°C for 24h prior to use. They were then crimped using aluminium bands over lids composed of a two-sided coating of polytetrafluoroethylene (PTFE) (inner) and butyl (outer) (SUPELCO SU860084, Merck, Darmstadt, Germany). The lids ensured that the sample was in contact with only glass or PTFE (inner surface of lid). The outer seal made from butyl 85 ensured air-tight re-sealing after sampling via a 0.7 mm needle. Subsequently, the vials were flushed with air containing equilibrated water vapour of known isotopic composition (hereafter referred to as "Source") for 10 min (see Fig. 1 for a photo of the setup) using the suction created by the cavity ring down spectrometer (CRDS, L2130-i, Picarro Inc., Santa Clara, CA, USA). Dry air was pulled from a laboratory gas drying unit (Drierite®, Fisher Scientific, UK), which dried the air down to 200 -500 ppmV according to the CRDS. The dry air supply was connected using a silicone tube forced over PTFE 90 tubing (1/4", Wolf Technik eK, Stuttgart, Germany) attached to a female luer-lock tube connector (CS -Chromatographie Service GmbH, Langerwehe, Germany) with an attached hollow needle (Henke-Ject®, 0.7x50mm, Henke Sass Wolf, Tuttlingen, Germany) on the other end. The connection between the source and the sample vial was similar, but with needles attached to both ends, while the final connection between the sample vial and the CRDS consisted of a needle on one end of the tube and a stainless-steel fitting (1/4" Swagelok, Stockholm, Sweden) on the connection to the CRDS (Fig. 1). 95 isotopic composition and the water vapour concentration are monitored for 10 min before the vial is disconnected from the flow and stored for later analysis.

100
We monitored the water vapour concentration and isotopic composition as we flushed the sample to be able to detect the time when the water vapour concentration stabilised. After stabilisation, we flushed the samples for two more minutes to allow for one more complete exchange of the sample volume, leading to a flushing time of 10 min in total.
We selected five sources of water with different isotopic composition to test this method not only for natural abundance applications but also for examining the applicability for labelling studies, where water enriched in 2 H is often used. Three of 105 the sources covered large parts of the natural abundance range for precipitation composition (i.e. "light", "medium" and "heavy"), and two more artificially enriched sources covered much of the labelled range (i.e. "very heavy" and "crazy heavy" see Table 1). The isotopic composition of these sources was measured on the CRDS using an autosampler and calibrating the measurements against "in-house standards" (δ 2 H: -102. where R is the isotope ratio of the sample or the known reference (Craig, 1961b).
Replicated vials were stored for 0, 1, 3, 4, 7, and 14 days, where storage of 0 days means the samples were analysed the same day they were collected ("0-day" samples). Samples were kept in racks at room temperature in the lab. Each source and 115 each storage time consisted of at least ten (five for the sources "heavy" and "very heavy") replicates. Before analysis, the racks with the samples were placed on a heating plate at 40°C for 10 min to reduce adsorption on the walls of the vials.
For sample analysis, the dry air supply and the CRDS were directly connected to the vial. We let the CRDS pull the sample vapour from the vial at the same time as dry air replaced the now missing volume in the vial (at ~35 ml min -1 ). This way the vapour concentration in the sample vial steadily decreased as the dry air diluted the water vapour. Because no water vapour 120 was being added, the isotopic composition of the sample remained unaffected (see scheme in Fig. 2). Again, the vapour concentration and isotopic composition were monitored.
We excluded the initial isotope purge by calculating the slope of the vapour concentration over time. We filtered out all data before f'(dH2O/dt) = minimum slope, which marks the beginning of the recession curve unaffected by ambient air and thus corresponds to the plateaus in the isotope data (Fig. 2). We then calculated the mean isotopic composition from the two 125 minutes starting starting from the time of the identified minimum slope (see yellow dots in Fig. 2B). We converted the vapour-phase measurements to liquid-phase data by assuming the vapour had been at equilibrium with the liquid water supply during sampling using Majoubes' fractionation factors (Majoube, 1971) and source temperature measured with a commercially available thermometer (TFA Dostmann 30-1012).

VSVS Field trial 135
We conducted our field test opportunistically during an ongoing tracer pulse-chase experiment. The pulse chase involved the addition of 2 H-enriched water (~1800‰ δ 2 H) to an area of approx. 200 m 2 surrounding a set of mature trees in a spruce-pine forest in northern Sweden. Briefly, we monitored the isotopic composition of the xylem water of eight tree individuals (four spruces and four pines) before and after application of the tracer for a total period of five weeks. We used the borehole equilibration approach as presented in Marshall et al. (2020). We drilled a 8 mm hole through each tree's stem, flushed it 140 with acetone to reduce pitch production and, after several days, connected the outlet side of the borehole to a valve unit, a pump and finally a CRDS to monitor the H2O concentration and isotopic composition. We refer to this setup with the term "in-situ system" from here on. The data presented here were collected on the last day of said experiment on September 1 st ,

2021.
We monitored a single scots pine (Pinus sylvestris) connected to the in-situ system. The selected tree was approx. 21.1 m high, had a diameter at breast height of 20.7 cm, and the borehole was installed ca. 40 cm aboveground, where the tree 145 diameter was 21.6 cm. Because this method has now been tested several times (Marshall et al., 2020;Kühnhammer et al., 2021), we used the calibrated in-situ data as our "true isotopic composition" of the trees' xylem. The calibration for the insitu data was conducted as described in Marshall et al. (2020). We then tested the new storage method against it.
Using the VSVS, samples were collected by connecting the "inlet" side of the borehole (in the original in-situ system this side was exposed to the atmosphere) to a gas-drying unit (Drierite®, Fisher Scientific, UK) and using a vacuum pump (no-150 name, 24 V, -50 kPa, https://www.ebay.de/itm/143587595483, last access 07/01/22) to draw saturated air from the "outlet" side of the borehole. In the original in-situ system this side was connected to the CRDS. A comparative scheme of the in-situ and VSVS setup can be found in the supporting information (Fig. S1). The pump was connected to a power regulator and a mass flow controller (MFC, MC-2SLPM-D/5M, MCS-2SLPM-D-.25NPT/5M; Alicat Scientific, Inc., Tucson, AZ, USA), both powered by a rechargeable lithium-ion battery (12 V, 12 Ah). This battery is suitable for use in a remote area as it 155 weighs less than 2 kg. With the present setup the pump and flow controller can run on the battery for more than 12 hours (see supplement for Excel spreadsheet with specifics). We set the flow rate of the MFC to 110 ml/min (which equals 77 µmol s -1 ) to match the flow rate used in the in-situ system. According to the modelling exercise in Marshall et al., (2020) isotopic equilibrium is reached using flow rates up to 150 µmol s -1 for trees of this diameter. As described in section 2.1 the vials were flushed for 10 min to allow the vial volume to be fully exchanged several times. The vials were filled sequentially 160 such that all vials for "0-day" storage time were filled first, then all for one-day storage time, and so forth. Standards (i.e. the sources "light", "heavy" and "very heavy" from the lab test) were prepared in the same way as in the lab test, with the modification of the higher flow rate and using the pump in the field. All standards and samples were assigned to a storage group (i.e. 0, 1, 3, 7, 14 days). All samples were stored in the lab until analysis, except for the "0-day" samples, which were measured directly in the field three hours after sampling. 165 Measurements were conducted as previously described in section 2.1, with the modification of measuring each sample for only 3.5 min. This was done because on the day of the field trial the inside of the borehole was colder than the lab during the lab trial. The sampled air was therefore less moist, leading to lower water vapour mixing ratios (wvmr, in ppmV) in the vials.
That meant the mixing with the dry air led to lower wvmr values more quickly than for the samples in the lab test, reducing the time period when wvmr were in the target range between ~17000 and 10000 ppmV H2O. This concentration range was 170 chosen to match the lab samples. We tried to avoid lower wvmr values as they generally associate with higher measurement uncertainties (https://www.picarro.com/products/l2130i_isotope_and_gas_concentration_analyzer, last access 07/01/22).
We switched from the battery-driven pump sampling to the in-situ system every four hours. As noted above, we compared the VSVS samples to the calibrated in-situ system data, which were considered our "gold standard". We disconnected the tree from the in-situ measurement system when not measured and re-connected it to the in-situ system 20 minutes before its 175 measurement was scheduled. In between we sampled the equilibrated vapour as described above (five per storage group).
We compared the "0-day" samples (n = 5) to the in-situ measurements of the same day (n = 2).

Analysis and Statistics
Calculations as well as graphical representations were conducted using the "tidyverse" packages in R (Wickham et al., 2019;180 R Core Team, 2020). To assess the VSVS's suitability to reliably store collected water vapour (assessing the "storage effect"), we calculated the change in isotopic composition (see Eq. 2 for either ∆δ 2 H or ∆δ 18 O) over the storage time (t), relative to the mean of the "0-day" sample (t=0) for each source and for the lab and field test, respectively (Eq.2): This "storage effect" was then related to the storage period using a linear regression model, separately for oxygen and 185 hydrogen as well as for natural abundance and enriched sources. The data were then corrected according to the storage period. We used the same model coefficients determined from the lab data to correct the field data samples. We additionally calculated the mean for each storage group (by source) and conducted pairwise Wilcox tests between the "0-day" samples and every other storage group, to disentangle effects introduced by the sampling method from storage. Wilcox tests were conducted using the "compare_means" function of the "ggpubr" package in R (Kassambara, 2020). 190 To relate measurements to the liquid true values we used a linear regression model for each storage group using the "lme4" package (Bates et al., 2015). We used three-point calibration for both δ 2 H and δ 18 O. That meant we separated the highly enriched sources from the natural abundance for δ 2 H, using the "heavy" source as the lowest standard for the enriched scale and as the highest for the natural abundance scale. The idea was to avoid "overweighting" the lower end of the enriched scale by adding three natural abundance standards to it. 195

Preceding work
The first tests for this method originate from a field trial in a boreal forest, where some of the authors attempted to trace an enriched water pulse through 120 trees simultaneously. Briefly, a hole was drilled through the entire diameter of a tree stem, equipped with brass fittings (Ahlsell AB, Sweden), and sealed from the atmosphere using chlorol-butyl septa (Exetainer, UK). Syringes (Henke Sass Wolf, Tuttlingen, Germany) were used to draw out 20 ml of equilibrated xylem sap vapour and 200 the isotopic composition was subsequently measured on a CRDS via injection into a dry air stream (Magh et al., 2021). The time between sampling and measurement varied between 20 minutes and up to 5 hours.
We noticed that the water concentration and isotopic composition of the vapour in the syringes were altered within hours after sampling. Though the test revealed suitability for heavy label detection studies where e.g., response times revealed by isotope dynamics rather than absolute values may be of prime interest. However, we do not recommend using plastic 205 syringes for long-term storage or for natural-abundance studies.  Table 1 shows the mean and the variation occurring immediately after the vials were filled ("0-day" samples). These data give an overview of the minimum possible variation (method precision) during the sampling procedure and compare it to the expected values defined through the measurement of the liquid source on the CRDS. Results depended on the source sampled (see sd values in Table 1), indicating that the vapour sampling procedure introduces higher variation than the liquid phase measurements (Table 1). 215

ID
d . Each storage group is compared to the 220 "0-day" sample group using a Wilcox-test (see supplemental Table S1 and S2 for δ 2 H and δ 18 O, respectively). The change in isotopic composition depends not only on the storage time but also on the enrichment in δ 2 H. The data show no consistent pattern regarding δ 2 H over storage times on the natural abundance range (Fig. 3 A). The median change ranges from 0 to 5‰ (Fig. 3A, Table S1). This observation is further supported by the linear regression model relating the change in δ 2 H to the storage period, as the slope of the fit is 0.06 and this model is not statistically significant (Fig. 4C), indicating that there was 225 no storage effect. However, for the sources enriched in δ 2 H the pattern reveals a constant depletion in δ 2 H over time. The median change for the enriched sources ranges from -10 to +17‰ for δ 2 H on storage day 1 and increases after that (Fig. 3B, Table S1). This is also described by the linear regression model fitting the change in isotopic composition over the storage period (R 2 =0.11, slope -3.67‰ day -1 , p≦0.05, Fig. 4C).  Looking at δ 18 O, the enrichment consistently increased with increasing storage period (Fig. 3 C and D, Table S1). The "storage effect" was well described by a linear model using the change in δ 18 O of all sources over the storage period. It is statistically significant and yields an R 2 of 0.43 (Fig. 4A).
The global meteoric water line reveals a tight relation between δ 2 H and δ 18 O with a linear fit and a slope of 8 (Craig, 1961a). 240 Thus, scales for the natural abundance sources in Fig. 3 were chosen to be eight times greater for δ 2 H than for δ 18 O to enable direct visual comparison of the storage influence on the composition. This was slightly bigger for δ 18 O as can be seen from Fig. 3, indicating less influence of storage on δ 2 H, which in turn, is also supported by the poor model fit (Fig. 4). We recorded the atmosphere's isotopic composition during sampling and the measurement days to check for admixture of 250 the atmosphere into the vial during storage (data not shown). We were thus able to rule out intrusion of atmosphere as all three standard sources would have been altered towards the atmospheric composition. This would have led to depletion rather than enrichment of heavy isotopes with increased storage periods, which was not the case ( Fig. 3 and Table S1). We then analysed the uncertainty of the stored vapour samples based on their true liquid isotopic composition. We used linear regression models for three natural abundance sources ("light", "medium", "heavy") and for three for the enriched 255 sources ("heavy", "very heavy", "crazy heavy") for δ 2 H, and all natural abundance sources for δ 18 O, at each storage time (Fig. 5). Overall, the model fits for δ 2 H are better than for δ 18 O, though both show high R 2 adj values. The high R 2 adj indicates that they are sufficient for empirical correction. The linear relationship between the liquid water and the measured vapour isotopic composition was statistically significant for all storage times (p < 0.01 for 10 samples per source and storage day), with similar slopes (Fig. 5). Though the slopes were similar we intended for the option to calibrate each set of samples with 260 their respective slope and intercept. The calibrated and uncalibrated data can be derived from Table S3 and are plotted in Fig. 6, showing that storage-effect correction and calibration reduces the variability between the storage groups, moving the samples close to their true liquid value. Fig. 6 Dual isotope plots of the raw mean (grey dots) and corrected/calibrated mean (coloured dots) of the lab trial storage data, corrected for the" storage effect" and calibrated using the linear regression models of each storage time. Sources are depicted by colour and the liquid true value is indicated by black stars. The upper panel shows the data of the two sources enriched in 2 H ("very heavy" and "crazy heavy", for the calibration we also used the "heavy" source, however we refrain from plotting it again here as it unnecessarily enlarges the Figure), while the lower panel depicts the three natural abundance sources (i.e. "light", 275 "medium" and "heavy").

Field trial results
The values of the VSVS samples were generally similar to the mean of the in-situ samples. The in-situ data revealed stable δ 18 O values (-13.15 ± 0.01‰) throughout the day, while in-situ δ 2 H varied up to 3.2‰ from a mean of 1.7‰ (Fig. 7). There were significant differences after some storage times as was indicated by Wilcox test (Fig. 7). The VSVS data failed to 280 return the in-situ δ 18 O when comparing the corrected and calibrated δ 18 O of the "0-day" sample to the in-situ measurements conducted on the same day (Fig. 7, Table S3). They became enriched relative to the source over longer storage times. In contrast, the VSVS δ 2 H data, as already observed in the lab trial, did not follow a constant pattern. VSVS samples stored for one and seven days did not differ significantly from the in-situ measurements (Fig. 7). In-situ data are depicted in black, while VSVS samples are indicated by colour. Wilcox test identified significant differences between the in-situ data and each storage group (differences are indicated by asterisks: "*" indicates p<=0.05 and "ns" not significant).

Time and Cost Efforts
To be able to make an informed decision about costs and time effort regarding the VSVS, we compare the VSVS to an in-290 situ system and destructive sampling and subsequent extraction via cryogenic extraction ( Table 2). The data for the latter two have been obtained from Kübert et al. (2020). Each method has its own advantages and disadvantages. In terms of equipment costs, the VSVS is the cheapest, even when including the running costs for repeatedly buying new lids and needles. In terms of time effort, the in-situ system and the VSVS are more efficient than obtaining and analysing samples for the cryogenic extraction line. Overall, the VSVS combines cost and time efficiency when compared to the two alternatives. 295 Table 2   We performed a lab trial of a water-vapour storage method using water sources covering stable isotope ratios in the natural 305 abundance range and well beyond it into the range highly enriched in 2 H. This was done to test the suitability of an in-situ approach to capture and reliably store water vapour combined with lab methods to analyse it. We added data from a field trial to further test the method's applicability under field conditions. Overall, we found the method to perform sufficiently well both in the lab and in the field within a defined range of precision and storage time.

Suitability of sampling method
We show that our adaptation of the in-situ method (Marshall et al., 2020) can simplify the analysis while reliably reproducing the isotopic composition of natural abundance samples when measured on the same day. The method is robust and cost efficient as it uses only a battery-powered pump and a flow controller to collect the samples, then the water vapour is stored in commercially available crimp vials, which can be re-used. 315 The reproducibility of measurements lies within the range reported for other in-situ approaches, e.g., Volkmann et al., (2016a). For example, the median reproducibility was 2.8‰ for δ 2 H and 0.33‰ for δ 18 O, while the uncertainty was up to 20‰ for δ 2 H and 3‰ for δ 18 O (Volkmann et al., 2016a;Beyer et al., 2020). In Marshall et al., (2020) the authors found their measurement precision to range from 2.3‰ to 7.8‰ for δ 2 H from natural abundance towards mild enrichment. For δ 18 O it ranged from 0.22‰ to 0.6‰. Thus, the VSVS provides a possible solution for settings where tree numbers are large, 320 sampling sites lie far apart, or laboratory facilities are at some distance. However, we point out that our sampling method did not reliably reproduce δ 18 O of the vials sampled in the field trial as it was about 1‰ more enriched than what the in-situ measurements suggested. We treat this result carefully as this was only the case for the samples but not the standards sampled with the same method in the field.
Although we were unable to reproduce the standard value within the above range for the samples highly enriched in 2 H (over 325 1500‰, "crazy heavy"), we do not regard this as surprising. High enrichment is generally associated with lower precision and samples outside the VSMOW-SLAP range cannot be calibrated to the same uncertainty level as samples within that range since in this case the requirement of "bracketing" samples with standards cannot be met. In labelling studies, the signal is usually so strong that higher noise can be tolerated.

Storage period significantly influences isotopic composition
In the crimp neck vials we observed a significant change in isotopic composition over time. The direction of change for δ 18 O was constant enrichment with longer storage time, indicating possible exchange with the atmosphere. The most reasonable explanation in this context is leakage through the lid from the higher water concentrations inside the vial towards the lower concentrations on the outside. However, this is a best guess scenario and the mechanistic studies of the changes of isotopic 335 composition in the vials were beyond the scope of this study.
Changes in δ 2 H were inconsistent and less pronounced than for δ 18 O. In most cases δ 2 H became significantly different after three days of storage, meaning that after three days the analytical range of variation was exceeded. For an overnight storage experiment with the "SWISS" system, Havranek et al., (2020) reported changes in isotopic composition between 0 and 1‰ for δ 18 O and between 0.3 and 4‰ for δ 2 H. Our data suggest changes between 0 and 0.5‰ for δ 18 O and between 0.4 and 6‰ 340 for δ 2 H when considering the natural abundance range. This indicates that for overnight storage both the VSVS and the SWISS perform on a similar level. However, when comparing our longest storage period (i.e. 14 days) to the 24-day storage in Havranek et al. (2020), it becomes clear that the VSVS does not sufficiently preserve the isotopic composition of its contents during the experiment, while the SWISS continued to perform accurately. For δ 2 H however, we found mean changes between 0 and 3.4‰ after 14 days for the natural abundance samples, which can be considered sufficiently small 345 depending on the research question. Nevertheless, we do recommend measuring samples within three days after sampling to get a result within the error margin of the methodologically introduced variation.
In addition to the smaller sensitivity of δ 2 H towards storage in the VSVS, we further recommend it for its low cost ( Table 2).
As noted above, all components are available off-the-shelf and, as of this writing, one litre of 99.9% 2 H2 16 O costs ~1000€, while one gram of 1 H2 18 O costs ~330€. 350

4.3
Isotopic changes with storage period can be corrected using linear models Given the potentially significant, yet systematic shift in VSVS data over time, we strongly recommend to prepare standards within the "equal treatment" framework as emphasised in e.g., Gralher et al., (2021). This means that the standards are sampled on the same day as the samples, stored under the same conditions and for the same period of time. One can then 355 presume that any systematic, storage time related isotopic shift in the samples is matched by the standards. Using this approach, we gained higher precision and accuracy for both lab-and field-based data.
For the field data set we emphasise the potential for additional variation due to the trees' water use and transport. As the sampled tree was constantly transpiring water throughout the sampling process and the sampling took roughly 50 min per storage group (for each sample, 10 min x 5 replications = 50 min), variations in the data may have originated from true 360 variation of xylem isotopic composition. Differences in the trees' xylem water isotopic composition over the course of day have previously been observed and described in, e.g., De Deurwaerder et al., (2020). In future studies this could potentially be avoided by reducing the sampling time per sample. The 10-minute sampling interval used here was derived from the low flow conditions of the CRDS in the lab trial, while the higher flow rate in the field trial would allow for shorter sampling times at the same sampling precision. 365

Conclusions
We introduced and tested a simple and cost-efficient approach to sample and store water vapour to enable plant or soil water isotope measurements that does not require access to line power. We proved the suitability of the sampling method within an extended precision range for natural abundance and samples heavily enriched in 2 H. We successfully tested the approach both in the lab and in the field. The isotopic composition of water was not significantly altered in a storage time of 3 days for 370 δ 18 O and δ 2 H was not altered beyond the variation introduced initially by sampling. This method extends the utility of in-situ sampling of water vapour, simplifying the collection and measurement of samples from which the isotopic composition of liquid water sources can be inferred.

Data Availability
Data are available from the corresponding author upon request.

Author contributions
RKM and JM designed the experiments. RKM and AK conducted the experiments. RKM and HL performed the data 380 analysis. RKM, BG, BH and JM conceived the theoretical parts. RKM wrote the first draft of the manuscript. All authors contributed with advice, prepared and reviewed the manuscript.