the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 02 Mar 2026
Combined and autonomous online measurement of water isotopes in precipitating snowflakes and atmospheric water vapor in East Antarctica
Abstract. Water isotopes in precipitation are a powerful tool to better understand the processes governing snowfall in Antarctica, which is essential to improve our knowledge of the Antarctic atmospheric water cycle and surface mass balance, and for the interpretation of past climate signals archived in ice cores. However, precipitation isotope observations in Antarctica rely on manual sampling, which is prone to fractionation under low accumulation rates and remains both time-consuming and logistically demanding, and thus restricted to stations and seasons where manual sampling is feasible.
In this study, we present a novel method that enables autonomous, continuous, and combined measurements of water vapor and precipitation δD, using a single laser spectrometer. This technique offers new observational capabilities to better capture the isotopic signature of snowfall events in polar environments. Compared to conventional manual sampling of precipitation, the technique is capable of analysing very small amounts of condensed water, while avoiding post-depositional effects. In addition, it enables high-temporal-resolution observations and is well suited for long-term deployments in unmanned environments. The sampling system prototype has been deployed at Dumont d’Urville station, located on the coastal margin of East Antarctica, and evaluated during three precipitation events from February to June 2023. A dedicated post-processing algorithm was developed to retrieve the isotopic composition of precipitation from the surrounding vapor background. Comparisons with independently collected snow samples show a mean deviation of -5.4 ‰ in δD, which is well below the observed intra-event signal amplitude of about 100 ‰. This demonstrates the reliability of both the sampling system and the retrieval algorithm to study the isotopic composition of precipitation at the event-scale. With this new dataset, two applications are proposed to better understand the water vapor – precipitation relationship at Dumont d’Urville: (1) an evaluation of the LMDZ6iso general circulation model, and (2) a comparison with ground-based remote sensing instruments (ceilometer and micro rain radar) to explore the potential of the Δ(δD) metric as a proxy for snow formation altitude. Beyond polar applications, the proposed method opens new possibilities for other types of observations, including liquid precipitation sampling or cloud water isotopes monitoring onboard aircraft.
- Preprint
(1299 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-256', Anonymous Referee #1, 27 Mar 2026
-
RC2: 'Comment on egusphere-2026-256', Anonymous Referee #2, 31 Mar 2026
In the manuscript titled “Combined and autonomous online measurement of water isotopes in precipitating snowflakes and atmospheric water vapor in East Antarctica” Thomas Lauwers et al. describe a new method to measure stable water isotopes of precipitation in coastal East Antarctica. Potentially this method could replace very labor intensive and unreliable manual sampling of precipitation while also improving time-resolution. With calibration and post-processing both water vapor and snow precipitation can be analysed quasi in-situ with one single instrument.
General comments
This new instrumentation together with post-processing is a great addition for process understanding of the water cycle. In Antarctica precipitation is known to be difficult to sample and to avoid fractionation or mixing with older snow or hoar. It is clearly pointed out to be a prototype and event-based study. Still, it would be helpful to discuss some points more comprehensively, e.g. by answering the following questions throughout the discussion:
- Could the instrument replace a dedicated instrument to measure stable water vapor isotopes?
- Could the instrument replace manual precipitation sampling?
- Is the instrument suitable for all kinds of precipitation (drifting snow, blowing snow, diamond dust, agglomerates)? What kind of “snowflakes” are expected to be able to be sucked into the sampling line with the status quo?
- What would be your recommendations for the optimal instrumentation?
To me, the measurement set-up is missing the option of a standard injection unit to calibrate precipitation measurements. At this point the comparison to manually sampled precipitation looks rather convincing, but I think a future version of the instrument would benefit a lot on the opportunity to independently calibrate the precipitation measurements. A syringe injection unit, like Affolter et al. (2014, https://doi.org/10.5194/cp-10-1291-2014) used, could help to optimise post-processing and design of the inlet line. Did you do any experiments to characterise sublimation in the snowflake line or liquid evaporation to optimise post processing?
In section 2 Methodology, I would prefer a subsection with a site description, where the climatic conditions (temperature, wind direction and speed, typical conditions for precipitation, drifting snow, blowing snow) the sampling sites (also in relation to prevailing wind direction) and also the remote sensing instruments can be described. It could be restructured a bit to make clear which instruments have been there before and what and when instruments were newly used for this study.
Specific comments
Line 35: Be careful with the word “first”, I would prefer “early”.
Line 36: Again “first”. Do you mean precipitation was measured before vapor? The title of Dansgaard, 1953, Tellus V is “The Abundance of O18 in Atmospheric Water and Water Vapour”, so cryotrap sampling was also used early on.
Line 78: Please replace “weird” with a more precise expression, maybe “extreme”.
Line 85: DDU abbreviation, please introduce every abbreviation first.
Line 91: See General comments. Please make a section with site description and introduce all abbreviations (e.g. also ATMOS). Add some details on the meteorology of the site and describe the remote sensing instruments.
Line 92: Which Picarro model is used?
Line 94: See above, what is the ATMOS building?
Line 96: I think it is low-humidity level generator.
Line 99: When was the new sampling line installed?
Figure 1/Right: All valve symbols except valve 6 have to be rotated by 90 °.
There are arrows drawn where the water vapor and snowflake inlet lines meet the sample lines of the analysers. I do not understand the meaning and would expect a simple T-connection.
Are there any open splits upstream of the analysers?
Line 109: What kind of filter (material and size)?
Line 112: What is meant with aperture? Is it just the end of the 1/8” tube? Is there a detail picture of the snowflake inlet available?
Line 112: What is meant with “to be pumped”?
Table 1: V6 is illustrated as a three-way valve, so it is not clear what open/closed means.
Line 125: What is the height of the sampling devices? Does the wind sock also collect drifting snow? Where is the sampling place in relation to wind obstacles?
Line 132: What were the criteria of event selection?
Line 149: This would mean that the Picarro would be a more suitable instrument to measure both stable water isotopes, at least for sufficiently high vapor backgrounds – later you mention that. Did you try the snowflake inlet with the Picarro at some point?
Line 176: In this set-up you can get the background humidity and isotopes from the Picarro also, or not? Does it change the results?
Line 180: That is why I think the set-up could benefit from the addition of a syringe standard injection unit or similar.
Line 183: Why the first 4 minutes? What are the criteria?
Line 186: Could you estimate the water content which would equal a 10 % increase in humidity?
Line 187: How common are situations like the last segment of figure 3? What could it be? How would figure 3 look like if you use the Picarro measurements as “background”, are there any differences?
Line 197: Please add a few more details for the events (maybe in a table): wind speed and direction, precipitation type and duration. Why did you choose these events? For the discussion: Are there additional events, where you doubt the quality of the results and what could be the reasons? Or are there other events where there was no accompanying high resolution manual sampling?
Line 208: Please introduce what subscript cp means.
Line 211 to 214: Move to discussion, but describe here the agreement/differences between the two sampling methods. Also, are there any obvious differences between the wind sock and tray samplings? Not everything you see in figure 4 is described. A definition of theoretical vapor is missing.
Figure 4: It should be avoided to use red and green in the same figure. Also red and violet are hard to distinguish. Please also indicate in the figure description what is blue and what is dark blue.
Line 230: Could you prove the kinetic effect in your sample line with some experiments? What kind of signal do you get outside of precipitation events, in calm conditions, during drifting snow, blowing snow events?
Lines 240 to 243 and figure 5: From the figure 5 it is not clear to me why an integration time of more than 4 min is required. The mean of the flake phase measurements seems still to be close to the collected samples, especially if you consider different sampling times as you mention in the next section? What would you expect the variability of snowflake water isotopes to be? How can you distinguish between measurement noise and “snowflake variability”?
Line 272: What are the conditions/the goal of the set-up for the discussion in section 4.1.1.? Do you assume you want to have one single instrument to get vapor and precipitation measurements at once and have no additional manual sampling of precipitation?
Line 291: Advantage over what?
Line 342 to 344: What would be the instrumental set-up for studies of snow formation, sublimation and surface interaction?
Lines 364 to 366: I do not understand.
Line 365: What is the CALVA program?
Line 366: Include ceilometer and micro rain radar in methods and explain what these measurements mean.
Line 385: You should mention the inverted axis already in line 379.
Line 429: How can the differentiation between drifting and blowing snow be done?
Technical corrections:
Figure 1: Snowflake inlet instead of Snowflakes inlet
Line 124: Snow sample collection and measurement instead of Snow samples collection and measurement
Line 143: Remove “situated”
Table 2: Dimensions of a1 and b1, the dot should be a multiplication symbol
Line 288: hours instead of hour
Line 378: on April 15th instead of in April 15th
Citation: https://doi.org/10.5194/egusphere-2026-256-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 188 | 72 | 15 | 275 | 11 | 11 |
- HTML: 188
- PDF: 72
- XML: 15
- Total: 275
- BibTeX: 11
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper addresses a significant logistical and scientific challenge in Antarctic climate research: the manual sampling of precipitation for stable water isotope analysis.
Traditional manual sampling presents logistic challenges and exposes samples to post-depositional effects (mainly sublimation and metamorphism) prior to collection, which can alter the snow pristine isotopic values. To overcome this, the authors present an autonomous method for the continuous and combined measurement of both water vapor and precipitation dD using a single laser spectrometer.
The aim is to provide high-temporal-resolution data to better understand the Antarctic atmospheric water cycle, evaluate general circulation models, and improve the interpretation of past climate signals in ice cores.
Specific comments
Line 21 (Abstract) I would not use the term “condensed water” here
Line 30 (Abstract) Is it LMDZ6iso or LMDZ6-iso (with the hyphen) as in other parts of the manuscript? I think it is LMDZ6iso
Line 52-53 change “Adding the water isotopes in the models permits to test and improve the representation of the atmospheric water cycle in the models” with “Adding water isotopes to the models allows for the testing and improvement of the representation of the atmospheric water cycle”
Line 78 Change “weird” with “anomalous”
Line 111-116 How do you differentiate between snowflakes and water vapor signal when using the “snowflake inlet”?
Figure 1 I see that the snow sampling tray is attached to a wood platform, which is higher than the tray itself, plus the sampler is very close to a blue shelter; don’t you think this could heavily interfere with the snow precipitation sampling?
The wind vane, which I suppose turns in line with wind direction, is also placed close to a building which shields the wind (and snow) from that direction
Line 181 The time of the plots seems to go between 15:30 and 16:30, more than between 15:00 and 18:30 UTC
Line 186 Change “In the example shown on Figure 2” with “In the example shown IN Figure 2”
Line 208 What is the “The δD of precipitation”? Does it refer to the snowflake sampling or to the manual precipitation sampling? It is not clear. You should also define δDp and δDcp
Line 228 Change “(Affolter et al., 2014) showed that differences” with “Affolter et al. (2014) showed that differences”
Line 294-295 While it is true that precipitation on the East Antarctic plateau has limited amounts, the frequency is not so low and you have to consider that it is also hard to discriminate between real snowfall events, blowing snow, hoar frost and diamond dust. All the aforementioned events produces snowflakes which, in your automated system, might be misinterpreted as snowfall, if not supported by other types of observations
Line 304-307 I think you have also to consider the input of possible blowing snow inside the inlet without precipitation occurring. A larger diameter inlet could facilitate the snowfall sampling, but it will eventually collect also more wind-drifted snow
Line 309-311 Compared to the system used in this study, the proposed setup would likely be more susceptible to blowing snow contamination
Line 335-336 Change “does not reproduce well” with “does not accurately reproduce” and change “simulations often showing a too smoothed evolution” with “simulations often showing an overly smoothed evolution”
Chapter 4.2.2 Your claim is that the metric Δ(δD) can be interpreted as an indicator of the temperature or altitude difference between the surface and the level of snow formation, provided that the vertical isotopic gradient remains approximately constant.
Are you sure that the vertical gradient remains constant? Isn’t it possible that precipitation occurs during temperature inversion and thus the vapor equilibrating with snowflakes is enriched in δD compared to the surface vapor you measured? Do you have vertical temperature profiles from the periods you studied? Have you also considered that snowflakes might experience sublimation during their descent?
Line 337-338 Although LMDZ6iso seems to capture pretty well the isotopic composition of precipitation, it looks like it fails to reproduce its variability within the event, especially for the final part of precipitation events; how do you explain that?
Figure 8 The caption should be clearer: Δ(δD) should be better explained and the difference between Δ(δDp) or Δ(δDcp) should be described
Line 375 Change “at 01:00 UTC” with “between 00:00 and 05:00 UTC”
Line 376-377 You have to specify that this rise occurs on April 17th
Line 378-379 You wrote “a first part occurring in April 15th before the cloud descent to ground level with low values of Δ(δD)”: first, change “in April 15th with ON April 15th”, then you say that the Δ(δD) values are low in this timeframe while, as confirmed in figure 8, and in line 381-382 you state that “In the first part of the event, from April 15th to before 03:00 UTC on April 16th, Δ(δD) shows much lower values, around ~-65 ‰ with a large variability. During this period, the MRR mostly (change “mostly with “mainly”) shows high reflectivity values (10-20 dBZ) extending from the surface up to 3 km, with cloud tops exceeding the radar’s observational range, indicating that snow particles likely originate from high altitudes.”
However, when looking at figure 8, the cloud base height during this timeframe seems quite low (at least after 05:00 UTC on April 15th); how do you explain it? Do you think this precipitation formed in a cumulonimbus (although, to my knowledge, these clouds were never observed at Dumont D’Urville Station, with the exception of pyrocumulonimbus clouds), at a significantly higher height than the cloud base?
Line 384-388 Please specify whether these two maxima in Δ(δD) are either Δ(δDp) or Δ(δDcp): looks like Δ(δDcp) to me. You should always specify when referring to either Δ(δDp) or Δ(δDcp), or to both of them
At some point in the paper you should talk about the snow accumulation from the three snowfall cases you studied and possibly compare the measured accumulation with the model output
Line 428-430 It is unclear whether this system can differentiate between drifting and falling snow. If it can, please detail the mechanism used to distinguish the two