The recent development of laser spectroscopy instruments now enables the continuous measurement of the isotopic composition of water vapor at many observation stations all around the world (Bailey et al., 2015; Bastrikov et al., 2014; Schmidt et al., 2010; Sodemann et al., 2017; Tremoy et al., 2011). In particular, the isotopic composition of water vapor has proven to be a very useful tool to document moist synoptic events at many locations (Bonne et al., 2014; Guilpart et al., 2017). In polar regions, the water vapor isotopic signal is not only useful for detecting the origin of moist air (Bréant et al., 2019; Kopec et al., 2014) but also for improving the interpretation of the isotopic composition of water in surface snow and ice core archives (Steen-Larsen et al., 2014). Indeed, exchanges occur after deposition between the surface snow and the water vapor, leading to modifications to the isotopic composition of the former and, hence, of the archived ice (Casado et al., 2016, 2018; Ritter et al., 2016).

Obtaining continuous and accurate measurements of the water vapor isotopic composition expressed on the VSMOW–SLAP (Vienna Standard Mean Ocean Water – Standard Light Antarctic Precipitation) scale measurements of the water vapor isotopic composition at Concordia station in central Antarctica is a key scientific challenge, as the deep ice core drilled there, EPICA Dome C, provides the oldest continuous water isotopic record expressed on the VSMOW–SLAP scale to date (Jouzel et al., 2007). Thus,it is a key reference for the study of past climate, and a correct interpretation of the isotopic record relies on the quantification of the transfer function between climate parameters and water isotopic composition in ice, which is itself influenced by exchanges with water vapor in the upper layers of the firn (Casado et al., 2018). Such knowledge is also of the uttermost importance for the interpretation of water isotope records from the new deep drilling project “Beyond EPICA – Oldest Ice” (https://www.beyondepica.eu/en/, last access: 5 April 2021), the aim of which is to drill a 1.5 million-year-old ice core at the Little Dome C site located 40 km from Concordia station and, hence, with similar low temperature and humidity conditions.

One of the main limitations of the current commercial instruments when deployed in polar regions is their relatively poor performance at low water vapor concentration. Generally, the precision of the measured isotopic ratios, δ18O and δD, rapidly worsens when the water mixing ratio decreases to humidity levels below 3000–5000 ppmv (parts per million by volume) (Bonne et al., 2014; Weng et al., 2020). However, in remote continental areas in Greenland and Antarctica, temperatures in winter can drop to very low values, leading to humidity levels down to 10 ppmv (Genthon et al., 2017). Arguably one of the most extreme experiments for continuous measurement of the water vapor isotopic composition was the deployment of a commercial Picarro L2130-i instrument at the East Antarctic French–Italian station of Concordia where the mean annual temperature is around -54 ∘C and the humidity barely exceeds 1000 ppmv during the warmest summer days (Casado et al., 2016). For such applications, there are two major impacts of low humidity on the raw isotopic signal: first, we generally observe an apparent increase in δ18O and δD with decreasing humidity level; second, the standard deviation associated with the continuous measurements of δ18O and δD of the water vapor increases. This can lead to overall uncertainties of several per mil for δ18O and tens of per mil for δD. Therefore, it is of uttermost importance to have a correct determination of the humidity dependency of the water vapor isotopic ratios.

Commercial instruments from Picarro Inc. are usually associated with a Picarro standards delivery module (SDM) designed to generate humidity at stable levels between 5000 and 30 000 ppmv. Using such a setup for humidity levels below 5000 ppmv leads to large uncertainties in the determination of the humidity influence on the water vapor isotopic composition (e.g., Guilpart et al., 2017). These uncertainties are due both to the instability of the water vapor generation using the SDM (in terms of water concentration – humidity – and/or isotopic composition) and to the analytical noise in the spectroscopy measurements when the absorption signals are weak. An alternative commercial device is the LGR (Los Gatos Research) calibration system (Water Vapor Isotope Standard Source, WVISS), which uses a nebulizer to instantaneously evaporate microdroplets of liquid water from a standard reservoir into a large (1 L) vaporizing chamber (Dong and Baer, 2010). This system is very stable and well adapted for a humidity range between 2500 and 25 000 ppmv (Aemisegger et al., 2012).

Several homemade water vapor injection systems have been developed with the specific aim of achieving better generated humidity stability at low humidity levels. A first approach is to use a dew point generator that injects small amounts of water into dry air (Lee et al., 2005; Wang et al., 2009). This approach is time-consuming, as it takes a long time to reach equilibrium and relies on a very precise knowledge of the temperature to quantify the isotopic fractionation. A method using a piezoelectric microdroplet generator into a dry-air stream could generate water mixing ratios between 12 and 3500 ppmv (Iannone et al., 2009; Sturm and Knohl, 2009; Sayres et al., 2009). However, adjustment of the humidity level and long-term stability are difficult to obtain with such devices. Systems relying on the use of syringe pumps have also been built by Gkinis et al. (2010) and Tremoy et al. (2011): a small fraction of the input stream of liquid water is introduced into a hot oven where water is vaporized in the presence of a dry-air flow. These systems cover a humidity range between 2000 and 30 000 ppmv. Finally, bubbler systems, in which dry air flows through a large volume of water to create saturated vapor, are very robust but can only produce water vapor at high humidity levels (Ellehoj et al., 2013). The aforementioned devices are unfortunately not well suited for automatic long-term operation at low humidity levels. During the 2014–2015 summer field season at Concordia station in Antarctica, a homemade humidity generator specifically designed for low humidity levels (Landsberg, 2014) was deployed (Casado et al., 2016). The device used dual high-precision, low-volume, syringe pumps to generate stable humidity levels at two different isotopic compositions over the range from 100 to 800 ppmv (Casado et al., 2016). Unfortunately, we observed quite a large scattering among the isotopic values measured at similar humidity levels as well as a large discrepancy between the humidity dependency of the water isotopic ratios measured in the field and that measured in the laboratory. Upon return to the laboratory, these defaults were traced primarily to tiny leaks in the water supply lines to the syringes.

Therefore, we reengineered the prototype by Landsberg (2014) in order to develop a robust and autonomous device for stable low-level humidity generation for the purpose of precise humidity calibration of spectroscopic instruments. Such devices have now been operating with minimum manual intervention for more than 1 year at two polar stations in Antarctica, Dumont d'Urville and Concordia, coupled to Picarro laser spectroscopy instruments. Here, we detail the technical description of the instrument and show key performance characteristics, enabling, for instance, a discussion of small-amplitude signals such as the diurnal variability of the water vapor isotopic composition in remote dry sites in East Antarctica.

The low humidity level generator (LHLG) developed here relies on the same principle as the one developed by Landsberg (2014), i.e., a steady, undersaturated evaporation of a liquid water droplet at the tip of a needle into a dry-air stream inside a small evaporation chamber. Based on this first prototype, the instrument has been remodeled including a specific hardware and software design.

The stability of the instrument has been tested over a large range of parameters. We show an example in Table 2. We modified the air flow associated with Standard A (the same results can be obtained with Standard B) between 200 and 400 sccm with an air flow on channel B of half the value of channel A. The infusion rate was varied between 0.03 and 0.14 µL min-1 in order to produce humidity levels of 400 and 800 ppmv. The 1σ standard deviations observed over 10 min plateaus are comparable to the standard deviation obtained when the air flow is set to 300 sccm.

For routine measurement, the air flow and infusion rate have been adjusted to optimize the stability of the generated vapor while minimizing the dry-air consumption. Thus, the LHLG is able to generate stable levels of humidity (drift lower than 20 ppmv over 1 h and 1σ below 10 ppmv over 10 min) from 70 to 2400 ppmv following the optimal set points shown in Table 3.

The main application of this device is the interpretation of water isotopic profiles at dry sites, in particular in polar regions. As shown in Fig. 5, the influence of humidity on the measurement of the water vapor isotopic composition with the L2130-i analyzer is large when humidity is below 1000 ppm and increases when humidity decreases. Even though the precise isotope ratio–humidity calibration curve is likely to be different from one analyzer to another, all laser-based water isotope analyzers investigated to date have shown a strongly nonlinear response at low humidity levels (Guilpart et al., 2017; Leroy-Dos Santos, 2020; Weng et al., 2020). At Concordia station, even in summer, humidity is generally below 1000 ppmv (Fig. 6); thus, the interpretation of the diurnal variability of the water vapor isotopic composition is strongly affected by the dependency of the measured δ18O and δD signals on humidity at this site. Figure 6 displays such diurnal variabilities during austral summer 2018–2019 at Concordia and the consequently large correction of the isotopic records (uncorrected in gray and corrected in black).

The data clearly demonstrate the importance of the humidity correction which generally shifts the curves to lower isotopic ratio values. However, the difference between uncorrected and corrected data is particularly important in the observation of the diurnal variability; this is better illustrated by zooming in on a section of the data, as in Fig. 7. When looking at the diurnal variability in the raw δ18O and δD isotope data in detail, some periods stand out with two identified daily peaks: one in phase with the humidity peak (marked in red in Fig. 7) and one occurring during the period of minimum humidity (marked in blue in Fig. 7). The strong nonlinearity of the calibration curve in Fig. 5 suggests that artificial peaks in δ18O and δD could be due to changing humidity levels. Indeed, after correcting the data for the humidity dependence of the analyzer (black curve in Fig. 7), the isotopic peaks occurring during humidity minima are diminished or disappear altogether, whereas the peaks occurring during humidity maxima are amplified. More strikingly, the phase of the signal changes by practically 180∘ over some periods. Whereas the raw isotope signal peaks during the night, the corrected record shows higher isotope ratios during daytime. The diurnal variability recorded on both raw and corrected isotopic values during a period with a higher humidity level – hence when the isotope ratio–humidity correction is smaller (around day 355 in Fig. 6) – also shows that the δ18O (δD) diurnal cycles are indeed in-phase with the humidity cycle. This result confirms the correlation between humidity cycles and δ18O and δD of the water vapor at the daily scale at Concordia, as reported by Casado et al. (2016). Thus, we conclude that the anticorrelation of δ18O and δD with humidity in the raw data (highlighted in blue in Fig. 7) during periods of low humidity is an artifact due to the influence of the humidity level on the vapor isotopic measurements by the L2130-i analyzer.

We have developed an autonomous instrument for low humidity generation (70 to 2400 ppmv) with controlled water vapor isotopic composition that is specifically aimed at carrying out continuous measurements of the water vapor isotopic composition using a laser-based spectrometer in regions characterized by very low humidity, such as polar regions. If needed, an interface permits the user to conveniently connect the new LHLG to the valve sequencer port of commercial Picarro instruments. After more than 1 year of routine operation at two Antarctic sites (Dumont d'Urville and Concordia), this instrument has proven to be very reliable and robust. It consistently generates stable humidity levels with a 1σ variability lower than 10 ppmv over more than 10 min. Moreover, its performance is significantly better than that of the Picarro SDM at low humidity.

We used this instrument for the calibration of our water isotopic data with a special focus on accurately quantifying the influence of humidity on the measured isotopic composition of the water vapor. This effect is huge at low humidity. We showed that this has an important impact on the interpretation of the diurnal cycles of δ18O and δD in the water vapor at Concordia station at humidity below 1500 ppmv. We were able to confirm that, at this site, the diurnal δ18O and δD variability is actually correlated with humidity variability, which would not have been possible without the new LHLG instrument.

Finally, the development of such an instrument is an important step forward to a better understanding of the transfer function between climate parameters and the isotopic composition of deep ice cores from the remote East Antarctic plateau, especially in the context of the new program “Beyond EPICA – Oldest Ice”. It should be completed by ongoing development of laser spectrometers better adapted to low humidity levels, such as those based on the technique of optical feedback cavity-enhanced absorption spectroscopy (OFCEAS) (Casado et al., 2016; Landsberg, 2014; Landsberg et al., 2014).