Continuous monitoring of surface water vapour isotopic compositions at Neumayer Station III, East Antarctica

In this study, the first fully-continuous monitoring of water vapour isotopic composition at Neumayer Station III, Antarctica, during the two-year period from February 2017 to January 2019 is presented. Seasonal and synoptic-scale variations of both stable water isotopes H 2 O and HDO are reported, and their link to variations of key meteorological variables are analysed. Changes in local temperature and humidity are the main drivers for the variability of δO and δD in vapour at Neumayer Station III, both on seasonal and shorter time scales. In contrast to the measured δO and δD variations, no 5 seasonal cycle in the Deuterium excess signal d–excess in vapour is detected. However, a rather high uncertainty of measured d–excess values especially in austral winter limits the confidence of this finding. Overall, the d–excess signal shows a stronger inverse correlation with humidity than with temperature, and this inverse correlation between d–excess and humidity is stronger for the cloudy-sky conditions than for clear-sky conditions during summertime. Back trajectory simulations performed with the FLEXPART model show that seasonal and synoptic variations of δO and δD in vapour coincide with changes in the 10 main sources of water vapour transported to Neumayer Station. In general, moisture transport pathways from the east lead to higher temperatures and more enriched δO values in vapour, while weather situations with southerly winds lead to lower temperatures and more depleted δO values. However, for several occasions, δO variations linked to wind direction changes were observed, which were not accompanied by a corresponding temperature change. Comparing isotopic compositions of water vapour at Neumayer Station III and snow samples taken in the vicinity of the station reveals almost identical slopes, both 15 for the δO–δD relation and for the temperature–δO relation.

the measurements in 1981 no significant temporal trend in air temperature has been observed (also shown by Medley et al., 2018).

Meteorological observations
In this study, temperature, humidity, wind speed, and wind direction data, measured 50 meters away from the station at 2-meter height above the surface, are used. We use meteorological observations at a resolution of 1 hour for the period February 2017-105 January 2019. To identify specific days with extreme high or low temperature, we determine multi-year daily temperature averages over the 38-year period from 1981 to 2018 (König-Langlo, 2017) and calculate the daily temperature anomaly at Neumayer Station. A day is considered as a warm (cold) event, when its temperature anomaly is higher (lower) than one standard deviation above (below) the mean.

Water vapour isotopic observations 110
In January 2017, a PICARRO L2140-i cavity ring-down spectroscopy analyser (simply named Picarro analyser, hereafter), has been installed in a lab room at Neumayer Station. The Picarro analyser is running continuously and measures the injected water vapour content and its isotopic composition approximately every two seconds. For our analyses, the data are merged to hourly mean values, if not stated otherwise.
For the largest time of the year, wind at Neumayer Station blows from easterly, southerly, or southwesterly directions. 115 Therefore, the air inlet for the vapour measurements is located on eastern side of the roof towards the mean wind direction and at its southern end to avoid exhaust gases. Apart from the time when the instrument is calibrated (see below), ambient air is constantly transported by an electrical pump from the rooftop to the Picarro analyser. During the calibration, to avoid keeping old air inside the tube, the pump continues sucking the air from the rooftop without sending it to the Picarro analyser. The whole inlet tube (approx. 10m long) is constantly heated to approx. 65 • C to avoid any condensation of vapour within the inlet 120 tube.
Part of the instrumental setup is a custom-made calibration system using three different isotopic water standards (liquid), with δ 18 O values of −6.07 ± 0.1 ‰, −25.33 ± 0.1 ‰, and −43.80 ± 0.1 ‰. δD values of the standards (water liquid) are −43.73 ± 1.5 ‰, −195.21 ± 1.5 ‰, and −344.57 ± 1.5 ‰. The chosen isotope values of the standards cover the whole range of expected isotope values in vapour at Neumayer Station (from -17 ‰ to -54 ‰ for δ 18 O and from -120 ‰ to -404 ‰ for δD) 125 during the course of a year. Isotope standards are provided in liquid form, and a bubbler system similar to the one described in Steen-Larsen et al. (2013) is used for vapourizing and measuring the standards. For safety reasons, two independent cooler boxes with two isotopic standards, each, are installed, and the isotope standard with a δ 18 O value of -25.33 ‰ is measured in both cooler boxes. The boxes are held at a constant temperature of 17 • C and air and water temperatures inside each bubbler system are constantly logged to determine the isotopic value of the vapour stemming from the liquid standards during the 130 calibration measurements.
For calibrating our isotope measurements, the calibration protocol developed by Steen-Larsen et al. (2013) and Bonne et al. (2014) has been applied and modified. The calibration procedure includes (i) the correction of isotope measurements at low 5 https://doi.org/10.5194/tc-2020-302 Preprint. Discussion started: 27 October 2020 c Author(s) 2020. CC BY 4.0 License. humidity by determining the required humidity-response functions of the Picarro analyser, (ii) corrections of a potential longterm drift of the instrument, (iii) the correction for an offset between measured and real isotope values, and (iv) filtering the 135 data of special events, -e.g. weather conditions with a potential contamination of our vapour measurements by the station's exhaust gases, or days with temperature stabilization problems in the cooler boxes.
The range of humidity defined for the Picarro analyser is 1000 to 50000 ppm. At Neumayer Station, relative humidity easily reaches values below 1000 ppm in the austral winter. For humidity values lower than 2000 ppm, the analyser shows systematic errors with biases of more than 1 ‰ for δ 18 O (Casado et al., 2016). To correct these systematic errors, we need 140 to assess humidity-response functions for our Picarro analyser. Humidity-response functions for all four isotope standards are determined once every year. Isotope values for absolute humidity ranging between 100ppm and 10000ppm are measured several times and a best-fitting curve (2 nd degree polynomial) is calculated. We do not find any change in the fitting curve between the different years of calibration. For determining and correcting an instrumental drift and offset, all isotope standards are measured every 25 hours. The measured isotope values are compared to the expected real standard isotope values for offset 145 correction and measured isotope values over a period of 14 days are considered for drift correction. Screening of the data for special events with anomalous vapour or isotope data are performed afterwards.
During the calibration procedure, water vapour is produced from the liquid isotope standard within the bubbler system. Over the course of a year, this might lead to a change in the isotopic composition of the liquid standards. To correct for a potential change in the standards, samples from all liquid standards are taken and measured yearly.

150
The uncertainty of measurements contains the accuracy and the precision on corrected measurements using error propagation method. The accuracy is calculated based on a calibration program, considering the instrumental drift, deviation from known isotopic values, and systematic error according to humidity response functions and the precision is based on taking average on measured data for 1-hour corrected data as the output of the calibration program. From the determined humidity-response functions and calibration procedure, we estimate the mean uncertainty of the Picarro isotope data over the whole observational 155 period as 0.45 ‰ for δ 18 O, 2.99 ‰ for δD, and 3.03 ‰ for d-excess values.

Moisture source diagnostics
To study the origin and transport paths of water vapour to Neumayer Station, the Lagrangian particle dispersion model FLEX-PART (Brioude et al., 2013) enhanced by a Lagrangian moisture source diagnostic (Sodemann et al., 2008) is used in this study. Meteorological data needed for the FLEXPART model are taken from the ERA-Interim reanalysis dataset (Dee et al.,160 2011), provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). Due to computational constraints, we restricted our FLEXPART analyses to the year 2017, only. Air parcels are traced backwards from the final destination (Neumayer Station) for 10 days, similar to the setup described by Sodemann et al. (2008). The moisture source diagnostic based on the Lagrangian back-trajectories provides values of "moisture uptake" (in mm day −1 ) on a 1 • × 1 • grid. This parameter represents the amount of moisture injected to the air masses within each grid cell, contributing to the humidity at Neumayer 165 Station.

Temperature and water vapour measurements
For the observational period 17 February 2017 until 22 January 2019, daily mean values of temperature, humidity, δ 18 O, δD, and d-excess have been determined (Fig. 2). There exist some data gaps for humidity and related isotope values for these 170 two years of measurements. Water vapour isotope data is missing at some days because of maintenance or reparation of the instrument, measuring humidity response functions, or due to the removal of data outliers related to instable measurements of the Picarro instrument. In total, daily vapour and isotope data exists for 600 out of 705 days (85 %).
Daily temperatures at Neumayer Station vary between approximately 0 • C in austral summer and -40 • C in austral winter.
Daily values of humidity vary between 0.06 g kg −1 (corresponding to approx. 100 ppm) and 3.75 g kg −1 (approx. 6000 ppm).  February 2018, small variations with a peak in autumn 2018 can be detected until January 2019. Due to the limited period of two years, the dataset is too short to determine if any seasonal cycle of d-excess exists in our measurements. This is even more true, as the uncertainty of the measured isotope values depends on the humidity amount, which is much lower at Neumayer Station in austral winter as compared to the summer season. Based on the determined mean uncertainty of the 1-hour Picarro 190 data (Chapter 2), we estimate the monthly average uncertainty for d-excess in austral winter as high as 4.5 ‰ (for July 2017) while it decreases during austral summer down to 1.9 ‰ (for November 2017, see Table 2). These measurements are performed in a distance of 50 meters from the main station building at a height of 2 meters. A 195 comparison of both datasets indicates that the Picarro analyser data are reliable ( Fig. 3). The correlation coefficient between these two totally independent humidity measurements is 0.97 (N = 12198, hourly values between 17 February 2017 and 22 January 2019). The relationship between these two series of humidity measurements (q P icarro = 1.5q meteorology + 0.08, standard error of the estimate = 0.0022 g kg −1 ) has been used for the calibration of the humidity values measured by the Picarro analyser.  q Picarro = 1.50q meteorology + 0.08, r=0.97 Hourly data, N=12198, , std=0.0022 g/kg humidities close to the surface due to hoar frost formation, which locally removes moisture from the atmosphere. The inlet of the Picarro instrument is situated approximately 17.5 m above the surface level of the station. As the station is placed on 205 a small artificial hill, this surface level is approx. 7.6 m higher than the surface level of the meteorological mast placed 50m besides the station building. Thus the total height difference between the Picarro inlet and the height of the meteorological humidity measurements is approximately 22 m. To test if contamination by exhaust gases could be another reason for this data mismatch, the wind direction was analysed for those hourly Picarro humidity values which are much higher than the corresponding humidity values measured by the meteorological station. Most of the outliers coincide with a wind direction 210 from the south (and a few from the east), which excludes the possibility that a contamination by exhaust gases is the reason for the unusually high Picarro humidity values.

Relationships between water vapour isotopes and local climate variables
Next, we analyse the relationship between δ 18 O, δD, temperature and humidity to determine the key meteorological variables controlling the isotope signals in vapour at Neumayer Station.  with snow. In summertime, when the air temperature can rise above 0 • C, the surface snow will reach its melting point and start 225 to melt. For the melting process, the incoming radiative energy is partly used for latent heat uptake, keeping the near-surface temperature close to the melting point. Thus, even though the atmosphere temperature might go above zero, the surface and 2 m temperature will stay almost constant close to 0°C. This phenomena can explain the detected cutoff at 0 • C of the 2 m temperature (Fig. 4) and might also partly explain the lower correlation coefficient between the 2 m temperature and δ 18 O in summer, as the latter is most likely controlled by upper air temperatures. following the correlation coefficient between temperature and δ 18 O. To assess these variations quantitatively, we calculate the correlation coefficient of the natural logarithmic specific humidity and δ 18 O for each season (based on daily data). The highest 235 correlation coefficient is found in spring r = 0.79) and the lowest one in summer (r = 0.57). Like the summer temperature values, the summer humidity values at Neumayer appear to be limited, with a maximum value of approx. 4 g kg −1 (Fig. 5).
The reason might be the same as for the observed temperature limit. Analysing different seasons of the year, a negative correlation between δ 18 O and d-excess is detected for spring (r = −0.47), summer (r = −0.51), and autumn (r = −0.37), but for winter, no correlation exists (Appendix, Fig. A2). This pattern can be 245 detected also for temperature-d-excess and humidity-d-excess relations (Appendix, Fig. B1 and B2). There is a negative correlation coefficient between temperature and d-excess for spring (r = −0.28), summer (r = −0.38), and autumn (r = −0.19), but in winter a weak positive correlation (r = 0.17) is noticed (Appendix, Fig. B1). Anti-correlations between the natural logarithmic humidity values and d-excess for spring (−0.29), summer (−0.43) and autumn (r = −0.21) are slightly stronger than the ones between temperature and d-excess. For winter, a weak positive correlation between natural logarithmic 250 humidity values and d-excess (r = 0.14) is found (Appendix, Fig. B2).

Moisture source uptake and vapour transport
For further understanding of the seasonal different relations between δ 18 O, δD, Deuterium excess, and the meteorological variables, we analyse potential seasonal differences in the main moisture uptake areas for vapour transported to Neumayer Station, as simulated by FLEXPART (Fig. 6). In spring, major moisture uptake and transport happens from oceanic areas 255 northwest of the station at high to mid-latitudes. In summer, most of the moisture uptake occurs in the coastal areas close to the station and in the South Atlantic Ocean, although in summer more humidity comes to the station. In autumn, moisture uptake occurs again mainly close to or east of the station, similar to summer. For winter, the moisture amount transported to the station is substantially less than in other seasons and comes from a wide area of the Southern Ocean, partly even from the Pacific.

Wind and pressure pattern 260
The origin of the air masses measured at Neumayer Station depends directly on the local wind, which is characterized by relatively high wind speeds, with an annual mean value of 8.7 m s −1 (standard deviation= 6.1 m s −1 ). Two main wind directions are found (Fig. 7). The prevailing wind direction is east, caused by the passage of cyclones north of the Antarctic coast in the circumpolar trough. For our observation period, the highest hourly wind speed value was 37.7 m s −1 from the east.

265
The second, less frequent, typical wind direction is south to southwest, caused by a mixture of weak katabatic and synoptic influence, when Neumayer Station is situated between a cyclone to the east and an anticyclone to the west.
Pure katabatic winds are rare and restricted to extended high-pressure systems. Because the Ekström ice shelf slope is gently upward to the south, katabatic wind remains below 10 m s −1 (south-north direction). In addition to the local wind at Neumayer Station, vapour transport to the station is also strongly controlled by the larger-270 scale pressure and wind pattern. For all seasons, generally there is a high-pressure system above the Antarctic continent and a low-pressure belt surrounding the coast, thus north of the station (Fig. 8). This pressure pattern puts the station on average at the southern edge of a low-pressure system, which leads to a cyclonic circulation that transports vapour from the Southern Atlantic to Neumayer Station from easterly directions. As the pressure pattern related to this circulation is weakened in summer, far-field transport of vapour to Neumayer Station is reduced as compared to other seasons (Fig. 6).  Atmospheric temperatures and humidity values are closely linked via the Clausius-Clapeyron equation. We have checked this link for the Neumayer Station data, and find indeed a high correlation between temperature and natural logarithmic humidity values (r = 0.99; Fig. 9). Thus, the high correlation between humidity and δ 18 O can be explained by the high correlation 290 of temperature with both δ 18 O and humidity. As isotopic fractionation is primarily controlled by temperature, we therefore estimate for the typical synoptic situation at Neumayer Station temperature fluctuations as the main driver for both changes in humidity and the δ 18 O signal in water vapour.

Wind
Next we analyse the impact of another important climatic factor, wind, on the isotope values in vapour at Neumayer Station. In winds and a wind speed higher than the daily averaged southerly wind of 4.56 m s −1 (12 out of 17 days), measured δ 18 O 305 values are lower than the predicted δ 18 O values. This means that southerly winds can transport water vapour with a more depleted δ 18 O composition to the station, and such transport might occur without any significant temperature change.
Next, we analyse wind days with extreme high or low temperatures. During the observation period, on 86 % of all days with a warm temperature event at Neumayer Station, the wind came from the east. Such wind conditions are usually a result of a low-pressure system north of the station. In such a situation, the weather at Neumayer Station is typically relatively warm with 310 high humidity (88% of days with a relative humidity higher than 90% coincide with wind from the east) and cloudiness (85% of cloudy days, means days with a total cloud amount more than 80%, coincide with this wind). In this wind pattern, relatively higher temperature and humidity leads to more enriched δ 18 O values. During days with extreme low temperatures, the wind typically comes from south to southwest, and winds are generally weak. This weather pattern occurs when a cyclone has moved eastward, so that the former low-pressure area is replaced by a high-pressure ridge. In such a situation, wind speeds decrease 315 and the wind direction changes from easterly to southerly and southwesterly. The weak katabatic winds are strengthened by the synoptically caused air flow and bring cold and dry air from the East Antarctic Plateau to Neumayer Station, usually dissolving the clouds. Lower temperature and humidity result in more depleted δ 18 O in water vapour coming to Neumayer Station. Wind patterns and their effect are summarized in Fig. 11.

Key controls on vapour d-excess
Changes of d-excess in vapour generally are supposed to reflect different climate conditions at the moisture source region (Merlivat and Jouzel, 1979;Pfahl and Sodemann, 2014). For Neumayer Station, the Deuterium excess in vapour is weakly anti-correlated with the 2 m temperature (r = −0.25). The correlation between d-excess and humidity is slightly higher (r = −0.40), but not very strong neither. The analyses of the wind pattern revealed that there are two main pathways for air parcels transported to Neumayer Station, they stem either from the east or from the south-southwest. The main differences between 325 these pathways are related to humidity (humid air from the ocean versus dry air from inland) and temperature (relatively warm air from the ocean versus cold air from inland). Using the back-trajectory analysis with the FLEXPART model, different areas of moisture uptaken, from locations close to the station to regions of the Southern Atlantic, could be identified.
In order to better understand the effect of different pathways and water moisture origins that control d-  (Fig. 12). The moisture corresponding to low daily d-excess values is either uptaken in coastal are as east of the station (this occurs mostly in summer) or northwest of it in the South Atlantic Ocean.

345
The moisture corresponding to high daily d-excess values is mostly uptaken from locations close to the station. .

Reliability of back trajectory calculation
Wind direction and wind speed are two of the main factors used in our back trajectory calculations to determine the origin of air parcels heading to Neumayer Station. To check the robustness of our results, we compare ERA-Interim data wind data, which is used in this study for FLEXPART simulations to identify moisture sources and transport pathways to Neumayer Station, with 350 meteorological observations from Neumayer Station. The comparison reveals that the ERA-Interim dataset reproduces wind direction and wind speed around Neumayer Station well for most days (Fig.13). However, for cold events, the ERA-Interim dataset has a bias with respect to the observed frequency of southerly winds. The main reason for this bias in the ERA-Interim data might be the low number of stations in Antarctica, which are used to generate the ERA-interim reanalysis dataset. Due to this bias in the ERA-Interim data, the simulated moisture uptake and vapour transport pathways during cold events, when 355 katabatic winds from the south occur at Neumayer Station, should be taken with caution.

Comparison of water stable isotope measurements in Antarctic vapour
To further assess the results of our water stable isotopes measurements in vapour at Neumayer Station, we look at comparable vapour measurements in Antarctica. To our knowledge, our study is the first one measuring water isotopes in vapour in Antarctica throughout the whole year. Other studies have been performed only during austral summer (mostly December and 360 January), for limited periods of 40 days or less (Ritter et al., 2016;Casado et al., 2016;Bréant et al., 2019). Thus, for the comparison of our data with these previous studies, we focus on our austral summers results (

Dumont d'Urville station in Adélie Land
Most comparable to our measurements is a recent study by Bréant et al. (2019), who have reported water isotopes in vapour from the Dumont d'Urville station in Adélie Land, which is also located at the Antarctic coast, however opposite to the Neumayer Station (Fig. 1). The average temperature at the Dumont d'Urville station during the study period was -0.5   (Ritter et al., 2016;Bréant et al., 2019). the east related to cyclonic weather pattern centered over the Southern Ocean play a major role (Medley et al., 2018).
The differences of the available summer data sets from these three locations (Neumayer Station,Dumont d'Urville,Kohnen 410 Station) are summarized in Table 2.

Air-snow interaction
Recently, several studies have reported an exchange of water isotopes between surface snow and the vapour above the surface for Greenland Madsen et al., 2019) and Antarctica (Casado et al., 2018). Such exchange might be compare our vapour measurements to isotope measurements of snow in the vicinity of Neumayer Station. At Neumayer Station, fresh snow has been sampled after major snowfall events since 1981. In contrast to locations in the interior of Antarctica, snowfall events in coastal areas of the Ekström Ice Shelf, where the station is located, are evenly distributed over the whole year (Helsen et al., 2005). Mean isotope values of these snow samples from Neumayer Station are -20.54 ‰ for δ 18 O and 153.25 ‰ for δD, for the period 1981 to 2000 (Schlosser et al., 2004). The slope between the 2 m temperature and δ 18 O in 420 the snow samples was determined as 0.57 ‰ • C −1 (r = 0.69) for the period 1981 to 2000 (Schlosser et al., 2004). isotopes are driven by changes in the water vapour isotopic compositions. In our study, the agreement in the T-δ 18 O slopes and also in the δ 18 O-δD slopes for water vapour and snow samples measured at Neumayer Station is remarkable. Thus, the isotopic exchange between water vapour and surface snow at this location should be further examined in more detail in future research studies. If surface snow isotopic compositions are derived by changes in water vapour isotopes, ice core water isotope 440 records might be interpreted as continuously recorded paleoclimate signals, even for periods without any precipitation.

Conclusions
In this study we analyse the first continuous measurements of water vapour isotopic composition at Neumayer Station, Antarctica, over a period of two entire years (February 2017 to January 2019). This unique data set makes it feasible to study not only summer conditions, but seasonal differences and variations of water vapour isotopes at an Antarctic station. Our measurements  how paleoclimate signals can be continuously recorded by water isotope variations in ice cores, even for periods without any precipitation. It has been suggested that a local diurnal cycle of sublimation and deposition could cause this isotope exchange between vapour and snow, but more detailed case studies and a combination of vapour and snow isotope data are required to better understand this process.
The performed moisture source diagnostics based on back-trajectory simulations show that the moisture origin for Neumayer 465 Station depends on the season. With the frontal zone moving north in summer, moisture uptake for Neumayer is closer to the coast in summer, whereas in spring and fall the moisture has its origin in a wider region reaching farther north. However, the presented back-trajectory simulations are least reliable for cold periods due to a strong bias in the ERA-Interim data set concerning wind speed and the frequency of southerly winds. This particularly affects our analysis based on back-trajectory simulations when cold, dry and thus depleted air is advected from the continent. Also, the very high variability of surface 470 pressure around Antarctica has to be considered, and a longer study period would be desirable in order to get more reliable results.
The Picarro measurements at Neumayer Station are currently being continued and supplemented by surface snow sampling.
They will be used for a longer-term study in the future, which should also help to confirm and support the results of this study in more detail. 30 https://doi.org/10.5194/tc-2020-302 Preprint. Discussion started: 27 October 2020 c Author(s) 2020. CC BY 4.0 License.