This work evaluates rain and water vapor isotope ratios collected during the monsoon period near Pune, India using a Below Cloud Interaction Model (BCIM). By tuning the boundary conditions of the model, the work concludes that 23% of the raindrop mass evaporates on average near Pune. The work broadens our observational constraints on rain evaporation, which is an important process that influences model climate sensitivity and storm organization and intensity. However, some of the methodology would benefit from additional clarification. Comments to that effect are detailed below. 

1. BCIM input, set up, assumptions, uncertainties:

To run the BCIM requires an assumption about the background water vapor isotopic profile. The paper tries using a) a Rayleigh distillation, b) a profile from GCM output, c) an average satellite profile from TES, and d) tuning the isotope ratios so that the BCIM rain and vapor isotope ratios match observation. Only the last attempt produces an agreeable result, and in large part because it was tuned to do so. Should we be concerned that assumptions about other factors—drop size distributions, for example—could actually be causing the model-observation discrepancy and that their effects are simply being masked in the tuning? Some of the answer appears in the Supplemental, and I strongly suggest that this material be included in the main manuscript instead. In fact, Figure S7 suggests that results vary more strongly with either drop size or RH than with assumptions about the background isotopic profile. More in depth discussion of assumptions and uncertainties would be helpful—particularly uncertainties in the rain evaporation percentage, as well as clarification on BCIM input methodology. 

For example,

• I’m unsure why TES from 2005-2007 was averaged over a 4x4 degree grid box to represent the profile at a single observation site in the year 2019. AIRS or IASI would have provided 2019 data with much greater frequency and thus created an opportunity to evaluate not just a mean profile but variations from that mean and their impacts on model output. 
• I’m also concerned that multi-order polynomials are used to interpolate the vertical profiles, when there are not a sufficient number of observations to constrain the number of coefficients.

Other places to clarify:

L 435  says, “drop diameter at the ground is provided as input,” but how does this work? Isn’t the drop size at altitude the initial input for the BCIM?

L 469 says, “appropriate interpolations were carried out.” The interpolation method should be specified.

L 479: “The procedure is discussed…” but where? Below?

L 500 says the constants were estimated “by interpolation”. What kind? Linear? More detail would be helpful.

L 540 talks about “acheiv[ing] a reasonable agreement” through tuning. But more specifics are needed. What makes something reasonable? Is there a physical basis?

L 557-8 talks about needing to increase the d18O and decrease the dD profiles, but how is this done? Uniformly with height? (This seems to be what the figure shows). Or only near the surface or top? Again, further clarification would help.

L 525 says “one possible explanation [for error in Runs 1 and 2]” might be due to a missed ET signature. But what about the fact that the Runs are using average profiles from a GCM and 4x4 degree satellite average as a proxy for a single convective location?

2. RH as the primary control:

The paper argues that RH is the primary control on raindrop evaporation because RH varies more than other factors like drop size diameter. But I worry that this is inferred from an absolute value comparison when what might be more relevant is to evaluate how the standardized values differ. One can do this by standardizing the predictors or by looking at partial coefficients of determination in the regression. Without that additional step, I’m not sure that this argument is well supported. Moreover, earlier in the work (e.g. L 392) there seems to be a stronger emphasis on the importance of drop size. Is there a reason that the argument (apparently) shifts?

3. Other minor comments: 

L 237 talks about 0.5 standard deviations: this is an unusual choice. Why not go with a more standard 1- or 2-sigma envelope?

Figure 2 talks about four regions marked by shading, but they are actually bounded by vertical lines. 

L 296 talks about d-excess increasing while d18O decreases. The evaporation actually causes the rain d-excess to decrease and the d18O to increase. So while the relationship is self-consistent, it might be switched for clarity.

L 322 talks about deep convective systems being “controlled by different microphysical processes.” But, different from what?

L 375 provides a specific range, but this is only approximate. No sample actually gets to -20 permil. Also, the word “respectively” is not needed after the parenthesis.

Figure 5 might benefit from marking the 15 “evaporation” samples so that one can pick them out more easily.

It would also help to clarify somewhere in the text that Figure 4 shows absolute value differences between rain and vapor while Figures 6 and 7 show the difference when the rain values are converted to vapor in equilibrium with the rain. It took me a while to understand why I was seeing different axes values.

L 630: “intimate relations” is not the right phrase for this context. Correlations?

Conclusion #4: I find this point confusing, but maybe I just need to reflect on it a bit more. It suggests that local water vapor supply cannot be important to sub-cloud moisture, even though the paper argues that rain evaporation fraction is significant. Is the issue that a lot of the raindrop mass is evaporated, but that total amount of water vapor yielded is actually quite small? 

Collision-coalsescence is not mentioned even though it is an important precipitation process that the BCIM neglects.