Before the experiment several preparatory measures were taken. The 250 mL infusion glass bottles were dried overnight in an oven at 60 °C to remove residues, then sealed with a butyl stopper and an aluminum cap secured with a 32 mm aluminum cap closure clip. The glass bottles were flushed with dry air for 5 min before sampling. After measuring the samples, the bottles were cleaned and dried again for 24 h at 60–80 °C for reuse. For the 500 mL Aluminum-zip bags with zippers (these bags were not reused because reusing sampling bags is impractical; it requires thorough and repeated preparation to remove isotopic memory effects and maintain accurate measurements, a process that is too labor-intensive for routine field or laboratory use (Herbstritt et al., 2023)), silicone was applied as a septum for the syringes and dried at room temperature for three days. The bags were then heated overnight at 50 °C to remove potential organic residues. The bags were tightly sealed by folding the top opening and securing it with folding clips, without using heat sealing, to prevent the emission of any compounds potentially affecting the isotope values. The 1 L FlexFoil sample bags were purged three times with dry air through the valves prior to sampling to minimize memory effects and allow reuse.

After the preparatory measures, the containers were filled with dry air from a cylinder regulated by a pressure regulator connected to a mass flow controller. The air was passed through PTFE tubing (outer diameter: 1/8′′) connected to a syringe using a Luer adapter. Fine syringes (0.80×40 mm) were used for sampling to minimize the size of the perforations in the septa. Consequently, five 250 mL infusion glass bottles were purged with the sample for 12 min at a flow rate of 100 mL min⁻¹, with equilibrium between the inlet and outlet of dry air established by an additional connection to the CRDS isotope analyzer. An open split was used to avoid overpressure in the analyzer. Five 500 mL Aluminum-zip bags were filled in six minutes under the same flow conditions by inserting the syringe through the silicone septum into the bag. For the 1 L FlexFoil sample bags, a 1/4′′ PTFE tube was attached. This tube was connected via a Swagelok connector to a second 1/8′′ tube, which carried the dry air until the bag was filled.

The sampling configuration and container types were identical to those described above. However, this experiment differed by storing saturated water vapor of a known isotopic composition within the containers. Each container type (250 mL infusion glass bottle, 1 L FlexFoil sample bag and 500 mL Aluminum-zip bag) was prepared in five replicates and stored for 6 h, 1 d, and 3 d at different temperatures: Fridge, Room, and Oven (4, 20–25, and 40 °C).

An initial measurement was performed by directly measuring the isotopic water vapor with the CRDS analyzer; these values were used as comparison values for the evaluation. The vapor flow rate was chosen at 100 mL min⁻¹ for all containers. Based on preliminary testing, a 12 min sampling time was selected for glass bottles to achieve optimal isotopic stability and minimize memory effects. Conversely, a shorter sampling time 6 min was sufficient for the bags to prevent overpressurization and subsequent leakage. The vapor was generated in a closed system (headspace) where dry air flowed through a glass bottle containing the known isotopic reference. This system included connections with syringes and PTFE tubing regulated by an electronic mass flow controller, and an open T-shaped split before entering the analyzer to prevent overpressure (Fig. 2A and B).

Experiments 1 and 2 were essential for selecting the optimal container. Six evaluation criteria were defined (Fig. 1) to ensure objective selection process. These were divided into critical (criteria 1–3) and non-critical (criteria 4–6) categories. Critical decision criteria were: (1) isotopic stability; the capability of the sampling container to maintain the isotopic signature without significant changes; (2) tightness; the prevention of exchange with atmospheric air; and (3) storage time; the storage time of a particular water vapor sample. Non-critical criteria were: (4) reusability of the storage container; a measure to sustainability and long-term cost reduction; (5) portability and transport resistance; a handling criteria including weight, volume, and risk of damage or loss of tightness during transport; and (6) cost, based on supplier prices in the European market. Each criterion was rated on a scale from 1 to 5 (1= poor, 5= excellent) and assigned a relative importance which we defined as follows: isotope stability (70 %), tightness (10 %), storage time (5 %), reusability (5 %), portability and transport (5 %), and cost (5 %). The sum of the weighted scores allowed for an objective comparison among the evaluated containers. The container with the highest total score from the combined decision criteria was considered the optimal choice and used in the detailed testing in phase 2.

As both sampling of water vapor and measurement on the laser spectroscope require puncturing of the septum with a syringe, we were interested in the effect of the syringe sizes responsible for the puncture – in particular the syringe diameter. Though theoretically sealing, use of a thick syringe creates a substantial hole in the septum which might cause intrusion of atmospheric air and effect isotopic stability. On the other hand, a syringe to thin might not allow for the desired flow rates. Despite being a very detailed aspect, we considered it worth testing. Two syringe sizes (0.80×40 mm, 80 mm, and 1.20×40 mm, 80 mm) were used to assess whether the puncture in the septum affected container tightness isotope values. Tests were conducted using 250 mL infusion glass bottles filled with dry air. After sampling with both syringes, the bottles containing air samples were stored under different temperatures and storage times. This evaluation identified whether syringe puncture had a negative effect on tightness or whether tightness depended mainly on the closure system (see Fig. 1). It also determined the most suitable syringe size for water vapor sampling.

The precision of the water vapor sampling method was evaluated at different sampling flow rates using only the containers previously identified as the most suitable. Three reference waters with known isotopic compositions were used in this experiment, namely ALBI1 (enriched in deuterium oxide), BS (intermediate isotopic composition), and KEI (light isotopic composition). Their isotope values are summarized in Table 1.

The water vapor samples were generated under controlled laboratory conditions (Fig. 1). Sampling occurred at four different flow rates: 35, 75, 100, and 125 mL min⁻¹. Each combination of reference water and flow rate was repeated three times. This experimental design allowed us to assess the uncertainty of the sampling method and to identify whether specific flow rates significantly affect the accuracy of the isotope values measured in water vapor. Before sampling, the 250 mL infusion glass bottles were cleaned by placing them in an oven at 60–80 °C for 24 h. After drying, the warm bottles were transferred to a dry, insulated aluminum box and allowed to cool under protected conditions to minimize exposure to ambient humidity and prevent condensation caused by rapid temperature changes. Before use, each bottle was flushed with a strong flow of dry air for 1 to 2 min to remove any residual from the interior. The bottles were then sealed by placing a butyl stopper on the opening, and sealing it with and aluminum crimp cap and crimping pliers to ensure a secure and gas-tight closure. During water vapor sampling, the temperature was monitored using a sensor (Sensirion AG, Sht4x Smart Gadget). Approximately 100 mL of the reference standard was placed in a 250 mL infusion glass bottle. During the sampling period, a dynamic equilibrium was established between the liquid water and the headspace vapor, allowing isotope exchange to approach the temperature-dependent equilibrium fractionation factor.

The sampling system was designed to pass a controlled stream of dry air through the bottle containing the reference water, generating water vapor that had equilibrated isotopically with the liquid water before flushing the sampling bottles. A synthetic dry air cylinder (20.5 vol % O2, Rest N2, KW-free) with a pressure regulator was connected to an electronic mass flow controller (PN 35828, Analyt MTC, Müllheim, Germany), allowing the sampling flow rate be precisely regulated. Connections were made with 1/8′′ PTFE tubing. The end connected to the mass flow controller used Swagelok connectors, while the other end had a Luer-lock adapter with a long needle inserted through the septum into the glass bottle. A second, shorter needle was inserted through the septum to transfer the generated water vapor to the sampling bottles.

The three replicate sampling bottles were connected in series in this setup, vapor generated from the reference water first passed through the first replicate bottle, then the second, and finally the third. The sampling was conducted for 35 min. The outlet of the third bottle was connected to the CRDS to monitor the vapor signal and obtain an initial reference measurement. Before connecting to the CRDS, an open-split configuration was used to prevent overpressure in the analytical system. The sampling system configuration is shown in Fig. 2C.

For each of the three reference waters, samples were prepared at each of the four sampling flow rates: 35, 75, 100, and 125 mL min⁻¹. Each reference-water and flow-rate combination was replicated three times, resulting in 36 samples per storage time (3 reference waters ×4 flow rates ×3 replicates). Across the four storage times, this resulted in 144 samples per temperature condition and 432 samples for the entire experiment.

Figure 3 shows the results for the tightness of the various containers. It indicates a clear increase in H₂O concentration (ppm) over time (= intrusion of atmospheric water vapor) in the dry air samples, depending on the type of container, storage time, temperature, and volume.

The 250 mL infusion glass bottles showed a substantial increase in H₂O concentration after 3 d of storage, reaching approximately ∼2200 ppm at 40 °C and about ∼1250 ppm under ambient conditions (20–25 °C), while values at 4 °C remained lower at around ∼480 ppm. The 500 mL Aluminum-zip bags showed a similar pattern, but with the strongest increase under ambient conditions (∼1750 ppm). Under extreme temperature conditions, such as the Oven (40 °C) and Fridge (4 °C), they showed comparable concentrations of ∼1000 ppm after 3 d. The 1 L FlexFoil sample bags showed the least variability, with values of approximately ∼100 and ∼600 ppm across all storage conditions and times, with no significant increase.

An isotopic reference was used to assess whether the isotopic composition remained stable after sampling in three types of containers and during storage at different temperatures and times. The results show a varying degree isotopic deviation of the measured values from the known values of the isotopic reference source. This deviation strongly depends on temperature, storage time and container type, and generally increases with longer storage time. In the 250 mL infusion glass bottles, the isotopic deviations in δ18O remained stable within a range of ±0.9‰ across all three storage times (0, 1, and 3 d) under all temperature conditions. Slight variations in δ2H were observed, particularly for samples stored in the Oven (+5‰) and Fridge (-5‰) after three days, while samples stored at Room temperature showed a stable deviation of ±0.5‰ at all storage times. Overall, the results show that analyzing of the samples until 24 h after sampling provides reliable values without deviations of -1‰ for δ18O and δ2H (+2.3‰ was observed exclusively in the Oven for δ2H; see Fig. 4).

In contrast, the 1 L FlexFoil sample bags showed δ18O deviations within ±1‰ for samples stored under Fridge and Oven conditions on days 0, 1, and 3. However, samples stored at Room temperature showed deviations of -1.35 ‰ on the first day and a +1.08‰ enrichment after three days. The deviations in δ2H were more pronounced from 0 to 1 d, with Room temperature showing deviations of -3‰, and values reaching up to ±8 ‰ after three days under Room and Fridge conditions. In the Oven, deviations remained close to +2‰, indicating a clear trend toward isotopic enrichment under these conditions. These results show that deviations are expected in this type of container from the first day of storage and tend to increase over time. Additionally, the fact that the deviations varied depending on the temperature conditions indicates that this material reacts differently to temperature changes, which significantly affects the isotopic signature of the water vapor (Fig. 4).

Finally, the 500 mL Aluminum-zip bags showed the largest increase in isotopic deviations for δ18O, reaching -3 ‰ under Fridge and Room conditions. In the Oven, the samples had values of -2.5‰ on days 0 and 1, followed by a slight enrichment to +1‰ on the third day. For δ2H, isotopic fractionation toward depletion was observed under all three temperature conditions, with deviations of up to -25‰ in the Room, -16‰ in the Fridge and -12‰ in the Oven (Fig. 4).

Figure 5 shows the results of the multicriteria analysis performed to determine the optimal container.

The 250 mL infusion glass bottles achieved the highest overall score of 4.35 (87 %) and were therefore the optimal container for water vapor sampling and “winner” of the testing. They were characterized by isotopic stability during the first days of storage (±1‰ deviation for both δ18O and δ2H; Fig. 4). Although they showed a greater increase in H₂O compared to the other containers (Fig. 3), their reusability, low cost in the European market, and reasonable transport stability make them a practical and cost-effective option for field sampling campaigns.

As a second option, the 1 L FlexFoil sample bag performed well in terms of tightness with a score of 3.75 (75 %) (Fig. 5), although it showed greater isotopic deviations than the glass bottles, especially under Room temperature conditions (Fig. 4). Its reusability is limited to a maximum of three cycles, and it is the most expensive container on the European market. Its main disadvantage is its portability, as the occupied volume increases when filled, making it difficult to transport and store large numbers of samples and increasing the risk of damage during handling.

Finally, the 500 mL Aluminum-zip bags received the lowest score of 1.62 (32.5 %). This poor performance was mainly due to their limited isotopic stability (Fig. 4) and the increase in H₂O observed during the tightness tests, especially after 3 d of storage. Although they offer the advantage of low cost, their reusability is practically limited to a single use. In addition, their portability and resistance to transportation are unfavorable, as the occupied volume makes handling multiple samples difficult and the material is prone to punctures.

Overall, glass bottles offer advantages that make them the preferable container. The criteria supporting this choice are particularly related to their ability to maintain a reliable isotopic signature during the first day of storage under various temperature conditions. Additional information and the rationale for each assessed criterion are provided in Table S2 in the Supplement.

The values obtained include the initial H₂O concentration, which was 200 to 350 ppm. The results show that the H₂O content in ppm increases progressively with storage time and is higher at higher temperatures. Samples stored at 40 °C reached the highest concentrations after 7 d, around ∼4000 ppm, followed by samples stored at 20–25 °C at approximately ∼2000 ppm, and finally samples stored at 4 °C, which had the lowest H₂O concentrations of about ∼700 ppm after 7 d.

The trends in H₂O concentration in the dry air samples in the 250 mL infusion glass bottles were consistent between the two syringe sizes, indicating that syringe size had no significant effect on the increase in H₂O concentration as a function of storage time and temperature, as shown in Fig. 6.

Water vapor sampling was conducted at four flow rates (35, 75, 100, and 125 mL min⁻¹). The H₂O concentration (ppm)measured during sampling with the CRDS, reached a plateau after approximately 25 min at 100 and 125 mL min⁻¹ and after 31 min at 75 mL min⁻¹, whereas, it did not fully stabilize within 35 min at 35 mL min⁻¹. A stable plateau was observed at 100 and 125 mL min⁻¹, but at 35 and 75 mL min⁻¹, the isotopic plateau developed more gradually over the sampling period, particularly for δ2H (Fig. S1 in the Supplement).

The stability of the water vapor samples was highly dependent on flow rate, with δ2H showing much larger fluctuations than δ18O. The 35 mL min⁻¹ rate showed the greatest variability, with δ2H interquartile range (IQR) of ±5.38 ‰ and outliers reaching ±15‰, this was mirrored in the δ18O by a noticeable shift toward negative deviations (±1‰). Optimal precision was found between 75 to 125 mL min⁻¹, where 75 % of δ2H values remained below ±1.5‰. In this range, the 75 mL min⁻¹ rate achieved the lowest median deviation (±0.18‰), suggesting that higher flows stabilize the vapor stream against ambient interference (Fig. 7A). However, isotopic integrity degraded proportionally with time, a trend visible in both isotopes but amplified in the δ2H signal. At 1 d, the system reached its peak stability with a δ2H IQR of, only ±2.04‰ and median near zero (±0.22‰). However, by day 7, the median deviation climbed to ±0.76‰ with a dispersion of ±3.76‰. Notably, at this 7 d mark, δ18O displayed significant negative outliers, confirming that long-term storage triggers a coupled fractionation process-likely due to vapor escape-that affects both isotopic signatures (Fig. 7B).

Contrary to our expectations, Room temperature storage yielded the highest precision (median: ±0.14‰; IQR: ±2.72‰ for δ2H). While the Oven samples showed slightly more scatter, the Fridge treatment was surprisingly the least stable. It produced the highest δ2H median deviation ±0.59‰ and the widest IQR ±3.38‰, with extreme outliers of ±15‰. This increased variability in the Fridge, visible in both δ2H and δ18O, suggests that cold conditions might interface with the internal equilibrium of the 250 mL glass bottle (Fig. 7C). Finally, the deviations for δ2H were strongly influenced by the isotopic weight of the standards. KEI was the most consistent (IQR: ±1.83‰), whereas HB showed the highest deviation values with ±2.03‰ and an IQR of ±3.68‰, indicating greater dispersion and lower stability compared to the other standards. However, ALBI1 had a larger spread with an IQR of ±3.16‰ and outliers of ±15‰. δ18O fluctuations up to ±1‰ indicating that heavier isotopes are more susceptible to memory effects (Fig. 7D).

The Mean Absolute Error analysis confirmed that the experimental conditions significantly affect the precision of water vapor isotope measurements, especially for δ2H. The smallest deviations between measured and target values occurred when flow rates of 75, 100, or 125 mL min⁻¹ were used for water vapor sampling and samples were stored for 6 h to 1 d at ambient conditions with temperatures between 20 and 25 °C. Under these conditions the best isotope values are obtained, with deviations of ±1.52 to ±1.89 ‰. However, at a flow rate of 35 mL min⁻¹, the deviations increase to 3.55 ‰, and if samples are stored for more than 1 d (e.g., 3 or 7 d), the deviations also increase. Similarly, very low (4 °C) or high (40 °C) temperatures result in larger deviations. For δ18O, the MAE indicates that deviations are small, with a precision of ±0.6‰ under all evaluated conditions, indicating that only δ2H is directly affected by variations in temperature, air flow, storage time and isotope concentration (Standard), as shown in Table 2.

The analysis of optimal conditions for water vapor extraction indicates that more than one flow rate may be suitable, depending on measurement duration and temperature conditions. For samples stored for 6 h, both 100 and 125 mL min⁻¹ produced small and consistent errors for δ2H (<±1.3‰) and δ18O (<±0.5‰), while 75 mL min⁻¹ was acceptable for δ18O but showed variability up to ±3‰ for δ2H and therefore can only be considered a partial alternative. After 1 d, the flow rate of 125 mL min⁻¹ was most optimal, yielding errors below ±0.5‰ for δ18O and ±0.3‰–1.5 ‰ for δ2H. However, 100 mL min⁻¹ may still be acceptable under certain conditions, provided the vapor samples are not exposed to extreme temperatures, as this could otherwise result in δ2H errors exceeding ±2‰. During prolonged storage (3 and 7 d), only the flow rate of 125 mL min⁻¹ remained within the range of ±1.5 to ±4 ‰, while the other flow rates showed greater variability, resulting in lower reliability. Therefore, if samples are analyzed within a short time (6 h), 100 mL min⁻¹ is the optimal flow rate, while 125 mL min⁻¹ provides greater stability under storage conditions of one day or longer (>1 d) (see Table 3).

Relative errors were calculated for the sampling flow rates considered optimal (75 to 125 mL min⁻¹), with data categorized into three storage duration groups (6 h, 1 d, and >1 d). Figure 8 shows that isotopic concentrations remained relatively stable for up to 24 h after sampling, with values generally below 4.5 % for δ18O and 2.5 % for δ2H across all analyzed flow rates. However, a distinct increase in relative error occurred when storage duration exceeded 1 d; this trend was particularly pronounced for δ18O, which reached a maximum relative error of 5.6 % at a flow rate of 100 mL min⁻¹. In contrast, δ2H showed greater stability over time, with relative errors remaining below 3 %. Although sampling flow rates influenced the magnitude of the error, storage duration was the primary factor determining isotopic variability.