The manuscript by Landais et al. presents an interesting 2-years dataset of continuous water vapour isotopic composition observations in Amsterdam Island. Such dataset is highly relevant for the atmospheric and water isotope community because it contributes filling the gap of water vapour observations in the southern Indian Ocean. The main output of manuscript is the explanation of atmospheric process driving some sharp depletion events of water vapor δ¹⁸O  occurred within the 2-years time frame. Such events break down the strong correlation between δ¹⁸O and mixing ratio, which is typical of marine and land sites. The authors selected 11 events, for which they propose near-equilibrium rain-vapour interaction coupled to atmospheric subsidence following a precipitation event. The proposed mechanism is supported by other data (e.g. GEM for vertical subsidence) and from model simulations (isotope-enabled GCMs and back trajectories). The manuscript also highlights the importance of spatial (and temporal) resolution in iso-GCMs, providing evidence on how the short-lived δ¹⁸O negatives peaks are completely canceled out in a coarse model grid. To this end, I fully agree that such kind of atmospheric events are a good benchmark for testing iso-GCMs.

I would like to compliment the authors for having prepared such a well-written paper and I strongly recommend the paper for publication after a very few adjustments/corrections. In general, the interpretation and the discussion of results is sounding and easy to follow. All figures are clear (see my only comment on Fig.4). 

I only have one comment/question about the interpretation of the results. Why d-excess has been (almost) left out of the discussion? The authors clearly state that during the depletion events both δ¹⁸O_v-q and  δ¹⁸O_v-dexcess correlations break down. But how d-excess signal looks like during the event? If the d-excess doesn't change much, it would provide support to the hypothesis of rain-vapor interaction close to equilibrium than to rain-evaporation and to atmospheric subsidence, since evaporation of raindrops and free tropospheric air are associated with high d-excess.

Minor comments:

L149 Please rearrange number of gasses (4) and monitoring years (3).

L187 STP conditions: 273.15 K

Figure 4 Including a legend and reducing the size of the caption could improve readability.