• General: For readers who are not familiar with feets and inches, it would be helpfull to systematically add between brackets the lengths in international units.
• Figure 3 caption: d and e are not described in the caption. Try something like “Mean attenuation times t95% for δ18O (a) and δD (b) and t63% for δ18O (d) and δD (e) and t3 ‰ for D-excess (c)”?
  
  Also in this figure, how were the error bars estimated? Explain this somewhere in the data analysis section?
• Section 4.2: I might be missing some basic elements to understand this section. Maybe giving a few more sentences of background or explanation would be useful:
  • aren't the tubing cylindrical? If so why aren't the surfaces and volumes linear with length?
  • “the shape of the isotopuc attenuation curves remained similar”: is it just the shape that remains similar? It looks like it's more than the shapes, the attenuation times remain similar as well, and this looks like the most important result.
  • l 327: “However...”: why does it contradict the hypothesis that the isotopic memory mainly comes from the analyzer cavity?
  • Why do the rates depend on the fraction of water adsorption sites that are out of equilibrium, rather than the number of sites? Could you give a simple equation (e.g. for the first-order kinetic reaction) that would allow readers not familiar with this literature to understand this paragraph?
• l 399: “. .” -> “.”

Dear Editor of Atmospheric Measurement Techniques,

The manuscript by Meyer & Welp is a study aimed to show how different tubing materials affect the water vapor isotopic signal propagation inside tubings. The authors tested different kind of tubings by forcing the experimental setup with large isotopic step changes without changing the water vapor mixing ratio.The authors then discuss the shape of rising/falling edges and the timing characteristics of the step change curves (lag, rising time, t₆₃, t₉₆ etc). The results show very similar characteristics for all the tubing materials tested, regardless of temperature (tested at ambient temperature and 60˚C). This study can be highly relevant for the water vapor isotope community, since there is no clear evidence/agreement on what type of tubing is best suited for high frequency atmospheric measurements of water vapor isotope composition. In general the paper is well written and enjoyable to read. Results and concepts are clearly presented and discussed. Figures are of good quality and easy to interpret. However, there are some aspect of the design of the study and choices that I believe the authors must explain/address before the paper is accepted for pubblication. In conclusion, the manuscript requires a major revision in my opinion.

Major comment #1: I am not an expert of OA-ICOS but usually such instruments are equipped with large optical cavity. I will assume an optical cavity volume of ~830 ccm, following Aemisegger et al. (2012) . This volume is ~1.5 times larger than the inner volume of the largest tested tubes (100 feet, ³/₁₆" ID).  Moreover, the flushing rate of the instrument is ¹/₃ of the flow rate in the tubings under test (in fast mode). Therefore, the experiment setup allow to spot only differences at very low frequencies. Indeed, all the high frequency components of the step change are dampened because of long average displacement of water molecules in the optical cavity. Therefore, the conclusion that all the tested tubing types are OK for water vapor analysis is valid only for low frequency analysis (e.g. hourly observations) but not for high-frequency analysis (e.g. flux, aircraft etc). Since I don't know the characteristics of the TWIA I might be wrong. In case the cavity volume is smaller, please do not consider this comment.

Major comment #2: The plot reported in the supplement (S2) shows an unusual increase of the Allan variance at short averaging time (>~60 seconds). If the water isotope source is stable (invariant isotope composition with time) and the measurement system is stable (measurement noise is mostly white, instrumental drift is small) the minima in the curves should be found at longer averaging time and the drift (the increase in the adev curve) should be smaller. See e.g. Fig.7 in Aemisegger et al. (2012) or Fig.3 in Jones et al. (2017). This suggests that the target value of the step change is not stable (i.e. the target isotope value is changing with time in the time frame of the analysis ~1 hour). This might be due to the change in isotope composition in the source of water vapor? The authors already identify the DPG as a potential source of isotope variability. A correction of the source isotope composition using Rayleigh distillation might be necessary (mentioned at L119-120). It is not clear how large the fractionation of the standard water was during the tests . 

Major comment #3: Please consider to change the step change into the impulse response by computing the derivative (see e.g. Jones et al., 2017, Steen-Larsen et al., 2014). This will let you to discuss how the signal is attenuated by e.g. fitting a normal distribution and looking at the standard deviation of the distribution, which is an indication of the average diplacement of molecules inside your measurement system. For water vapor stable isotope analysis usually the impulses are not symmetrical, therefore a best fit of a log-normal distribution or of an exponentially modified Gaussian distribution should to the job. 

Minor comments:

- It is not clear how the start of the step change is detected. I think the swithcing of the 3-W valve is logged but how you detect the "start" of rising-falling edge to precise measure the lag?

- A spectral analysis of the impulse response could be beneficial for understanding the limits of each tubing material for each application (e.g. by identifying the 3dB attenuation and the passband)

- L185 does this means that the impulse response of your system is guassian? Or at least, symmetrical?

- L344 Fairly slow? In respect to stable isotope analysis?

- L494 Link to code/data is not working

References

Aemisegger, F., Sturm, P., Graf, P., Sodemann, H., Pfahl, S., Knohl, A., & Wernli, H. (2012). Measuring variations of δ 18O and δ 2H in atmospheric water vapour using two commercial laser-based spectrometers: An instrument characterisation study. _Atmospheric Measurement Techniques_, _5_(7), 1491–1511. https://doi.org/10.5194/amt-5-1491-2012

Jones, T. R., White, J. W. C., Steig, E. J., Vaughn, B. H., Morris, V., Gkinis, V., Markle, B. R., & Schoenemann, S. W. (2017). Improved methodologies for continuous-flow analysis of stable water isotopes in ice cores. _Atmospheric Measurement Techniques_, _10_(2), 617–632. https://doi.org/10.5194/amt-10-617-2017

Steen-Larsen, H. C., Sveinbjörnsdottir, A. E., Peters, A. J., Masson-Delmotte, V., Guishard, M. P., Hsiao, G., Jouzel, J., Noone, D., Warren, J. K., & White, J. W. C. (2014). Climatic controls on water vapor deuterium excess in the marine boundary layer of the North Atlantic based on 500 days of in situ, continuous measurements. _Atmospheric Chemistry and Physics_, _14_(15), 7741–7756. https://doi.org/10.5194/acp-14-7741-2014

The manuscript of Meyer and Welp details a comparison experiment of tubing types at two temperatures. The work deals with the common issue of memory effect in water isotope analysis and aims to minimize this effect by finding the most appropriate material. While in general the experimental setup is logical and the text reads well, I have some major points that I feel are not addressed well. Also, the text and figures need refining to more clearly communicate the findings.

Major comments:

• My main concern about this paper is highlighted in figure 4, where fast and slow analyser flow modes are compared. Based on the text, fast analyser mode increases the flow of air through the optical cavity (x2.5). During fast analyser flow, the MFM was removed, but test tubing was kept. Also, flow rates upstream of the analyser before the venting T, passing through the tested tubing itself, remained constant. Given that your study was designed to test tubing attenuation, and nothing changed in the tubing or flow through the tubing, no difference would be expected between slow and fast analyser modes. Still, Figure 4 indicates a 10x smaller memory effect duration for fast analyser flow compared to slow analyser flow.
  • This suggests that all (equally large) attenuation times found in slow analyser mode, the mode in which all tubing was tested, were predominantly caused by attenuation in the instrument or, as the authors suggest, in the MFM. Thus, not by tubing itself. How can reliable conclusions be drawn on tubing material type then?
• I would highly recommend including dekabon as an additional tubing material. In the introduction you clarify that it is known that dekabon causes attenuation. If you can show it also does using your setup, you can be more confident about the attenuation times you find for the other materials. In the current state, open questions about the setup cast doubt on your finding that attenuation times are independent on tubing material and temperature.
• Intuitively, tube length and tube ID impact attenuation, as indicated by your measurement setup. You don’t observe this and defend your findings claiming the exchange is a “first-order kinetic reaction” (Sect 4.2). I miss a simple, interpretable, explanation on this, given the context of tube attenuation, possibly including a figure. Please clarify in your explanation why the following train of thought would be incorrect:
  1. with 20x longer tubing, 20x more exchange sites are present, all occupied by isotopic composition 1 just before the switch.
  2. after the switch to isotopic composition 2, exchange sites are swapped with a constant rate for each isotopologue, independent of tubing length.
  3. as the longer tube has a 20x higher net amount of isotopic composition 1, it will take longer for the output signal to consist of 95% isotopic composition 2.
• You indicate in your introduction (e.g. L. 60) that temperature and air flow rate (and tubing material) have a known “great effect” on attenuation based on various previous studies. Still, your results replicate none of these effects (effective air flow rates through tubing material was changed by wall thickness variations in your experiment). I feel that section 4.1 and 4.2 don’t currently provide convincing arguments for why you don’t find the known dependencies with temperature or flowrate.

Minor comments:

Graphics

Fig 1.    Please add the flowrate coming from the WVISS. Is this exactly 1.1L/min? Otherwise, why doesn’t it need an overblow? Also, given that only mass flow meters (not controllers) are used according to the scheme and text, how is the 1.1 L/min set. Explain in the text if the rotameters were used to set the flows.

Fig 2.   Add a theoretical e-folding time based on sample cell mixing. I derive a 25sec residence time based on a 500ml cell (which I think your LGR has) with 100ml/min flow. It shows the reader whether analyser mixing can cause attenuation (seems non-dominant) but generates questions on how the analyser flowrate adjustment has such a large effect.

Fig 4.    This figure should be remade. The panel labels are in the wrong order compared to the description. The y-label of current panel b is wrong (delta D). The legend is unclear as it looks like one large legend while each column has its own.

Tab S1. Tubing lengths are only occasionally mentioned, include this everywhere and make units uniform (foot or ‘). Also, I notice that the WVISS dilution setting was not constant, seemingly affecting the H2O concentration of the mixture generated, why wasn’t the dilution constant?

General

• I noticed frequent incoherent sentence structures, sometimes making it challenging to get the point. I recommend going over the document to improve this.
• You are in a low flow laminar regime (Re << 3000) through a long tube, yet within-tube flow rate differences, which are roughness dependent, were not explicitly considered as a cause for signal attenuation. I think it would be a valuable point to add (even if effects are non-dominant).

Text

L.104    “fast” and “slow” analyser suggests you used multiple, which I understand was not the case. Also, it would help the reader to explicitly state whether the flowrate through the tested tubing changed (I understand it didn’t ) under both analyser regimes. Lastly, why multiple flow speeds through the analyser and not only maximal instrument flow? Increasing instrument attenuation complicates your setup not helping to answer your research question.

L. 108    Indicate in this section if any calibrations were performed and if not, why. This should also clarify whether the isotopic compositions you mention are raw analyser outputs or independent compositions.

L. 124    If possible, indicate a range instead of the “aproximately” as the consistency of the 100 foot length seems essential for your tests! E.g. +/- 10 foot or 0.1 foot?

L. 134    Clarify how the temperature was measured and guaranteed? In case 60C was the maximum heating temperature of the self-regulating heating tape, real temperatures could have been much lower (often a linear wattage decrease with no heat emitted at 60C, and hardly any at 50C. Given a heat loss through the insulation, a Win == Wout at 40C might also have been realistic).

L. 138    You mention multiple errors for the 1.1L/min flowrate. Is it 0.15, 0.45, or the combination? Relatively, errors seem large given your dependence on a consistent setup.

L. 148    Explain why the filter was needed. The dry air source and dryrite are likely already filtered, and standards used were likely demineralized. It seems like an extra uncertainty that is not evidently needed.

L. 209    Be more explicit about the nature of this breakpoint analysis. Is it the time from switch to any “new isotopic signal” hitting the analyser?

L. 248    The contents of this paragraph are near identical to the contents of the paragraph before it in another wording, consider merging both.

L. 323    The “shape” of the attenuation curve was not expected to differ, but the attenuation time is expected to differ. Remove “shape” in the text to prevent this confusion for the reader.

L. 327    If instrument influence is “likely” much larger than the tubing, the paper loses its merit, and the conclusions can’t be made. If this is indeed a real concern this discussion point should be expanded, or experiments should be repeated.

L. 350    Unclear argument. Was the flowrate adjusted in the “with omega / without omega” experiment in the appendix? It seems like it was, making it odd to say the omega was the cause of the attenuation, and not increased analyser flow.

L. 355    The arguments presented for MFM attenuation suggest that material type and additional volume are key, seemingly contradicting the presented conclusions stating that neither tube length (i.e. volume) nor material type impact attenuation. Please attempt to reconcile why this could be.

L. 369    Mention this residence (or turnover) time in the methods together with your instrument and flow details. Also, the 8-12 seconds seems to be based on the 0.2-0.3 flowrates while 0.1L/min was used for most tests.

L. 370    Unclear scentence. Define “this” (2x). you seem to suggest in-line elements impact the analyzers turnover time, but the analyser regulates its own inlet speed, correct?

L. 388    It is not entirely clear how you defined your lag time using the breakpoints. The unexplained lag time is similar to the order of magnitude of the sample cell residence time I found for 0.1l/min flow (25s).

L. 445    The recommendation to use short inlet tubing seems to contradict your own findings that inlet tube length does not matter for isotope attenuation times. Clarify that recommendation if you chose to keep it.

L. 475    Your experiment with reduced tube ID effectively increased the air flow rate in the test tube and you found no difference in attenuation whatsoever. Stating in your conclusion that “higher air flow rates will minimize the memory effect” seems opposed to this. Please explain this better earlier in your text or adjust the sentence.

L. 482    To keep the conclusion short, remove “While differences…” as this sentence has no different message than the sentence before.