I am excited to see this manuscript outlining current developments of CFA systems for snow cores being published.

Best, Sonja Wahl

This manuscript focuses on the continuous measurements of water isotopes in firn using a continuous flow analytical (CFA) set-up and uses this set-up, combined with disctrete analyses, to address the effect of diffusion of water isotopes during storage. The study presents some new measurements from 1 m sections (2.4 to 3.4 m) on 6 firn cores. Because CFA is affected to some mixing due to percolation, analyses of discrete samples performed in 2015 and 2019 over the same sections are also used to infer the effect of diffusion over 4 years. The authors deduce a diffusion length of 45 mm for storage diffusion during 4 years.

The manuscript is short, useful and in general well written. I suggest to accept it but I would like to suggest a few points of the study which should be explained in more details before acceptation.

• It was not clear to me what was new in the set-up presented here compared to the previous one in addition to the change of the melting head. It could be better explained.
• Is there any comparison between the performances of the old and the new melt-head for firn analysis and how could the improvement be quantified ?
• Why was only the section between 2.4 and 3.4 m studied ? Why not studying the effect of diffusion during storage at different densities (e.g. a section of very low density on the section covering 1 m at the very top of the firn, a section of 1 m at 3 m depth, a section of 1 m at 10 m depth and a section of 1 m near the close-off) ?
• Is it possible to make some recommendations from this study on the storage conditions ? For example, depending on the accumulation rate at each site, the diffusion may affect the recording of the seasonal signal. Is it possible to say that below a certain accumulation rate, the seasonal signal is no more visible after « a certain number » (to be precised) of years of storage at -20°C ?

    Then, I have some minor remarks :

- l. 34-37 : can you precise at which timescales the ice cores of the East Antarctic plateau are dominated by noise ? As such, without any quantitative indication, these sentences are not very useful.

- Figure 4 : to see the (small) deviations of each data point from the regression line, it could be nice to show on top of each figure the difference in d18O for each standards between the real value and the value calculated from the measured value and the regression line.

- Figure 6 : Is it possible to display an envelop showing the spread of the CFA profiles in addition to the stack ?

- In general, I was wondering why you chose a value of 22 mm for the discrete sampling. It is probably not very convenient for the sampling of 1 m core section. Why not 20 mm ? What additionnal information (if any) could we learnt with a higher resolution discrete sampling ?

- l. 319 : « Vostok » and not « Vostock »

- I did not get exactly what should be improved on l. 324. Is it possible to further explain ?