This paper addresses the long standing (and still of paramount importance) topic of convective cloud formation/development and the surrounding water vapour (WV) conditions that lead to it.

The problem of WV retrieval above clouds, although attempted in the literature (as mentioned by the authors), remains very appealing and very topical nowadays, because no systematic/operational retrieval exists at the moment in Meteorological agencies. The dedicated WV imager on the C3IL mission will cover this gap, allowing a 3-year long coverage.

The problem of WV retrieval above clouds is successfully addressed by proposing an optimal estimation approach, using SWIR imaging measurements in 3 relevant channels (covering the spectral range from ~1 to 1.3µm) in and off the water vapour absorption band and taking into account the relevant factors affecting (together with the water vapour profile) the observed radiances in these channels, i.e., the surface properties  such as albedo and the cloud optical thickness (COT) and height).

The retrieval approach is motivated based on previous works conducted with POLDER and MERIS data and demonstrated using appropriate simulated data, and test retrieval using both idealized and realistic atmospheric profiles. The results are convincing (i.e., the proposed algorithm is clearly sensitive to the water vapour amount above clouds and its quantification is reliable within well motivated uncertainty (i.e. impact of realistic cloud profiles, of high COT, of low WV content, etc.).

Although demonstrated on simulated data only, the method opens interesting perspectives for the 3-years limited C3IL mission, and maybe for possible longer future satellite missions equipped with the same channels as the C3IL/WV imager.

All this considered, I have no doubt in recommending the publication of this paper after minor revision. In the attached PDF, I added all my comments to the text. Most of them are typos corrections or request for clarifications (e.g., clarify in abstract/conclusion that the proposed algorithm is demonstrated only over ocean  and excludes latitudes higher the ±60 deg). However, I also added suggestions and comments that in my opinion may further improve the quality of this work (e.g., future work using real profiles and 3D cloud reconstruction from EarthCARE). The authors shall go through them, providing feedback where required.

This manuscript describes a retrieval algorithm of integrated water vapor content above cloud form shortwave infrared observations as will be measured with the C3IEL mission. The manuscript is of interest for AMT. However, I recommend some clarifications to be added to the manuscript, related to the questions below. Please address these issues in the revised manuscript.   

1. The relative humidity is assumed to be 100% in cloud. I assume here the relative humidity is defined with respect to liquid water. This assumption is then true for water clouds, but not necessarily true for ice clouds. What is the impact of this assumption for ice clouds?
2. For clouds with tops higher than 4 km, cloud tops are assumed to be ice. Many supercooled liquid clouds extend much higher than 4 km. What is the impact of a wrong cloud top phase assumption to the results?
3. What is the sensitivity of the results to the assumed base height? The sensitivity is briefly discussed in section 5.2.2., but I think it should be systematically investigated related to the analysis shown in figures 3 and 4.
4. In line 235 it is stated that “Consequently, less radiation is absorbed within the cloud with the SAS model used for the retrieval than with the AFGL tropical model used to simulate the test measurements.” This seems to contradict the statement in line 215 that “different first guess profiles, both in idealized and realistic scenarios, give similar results.” What is the real influence of the assumed profile?

Some minor edits:

Line 34: emits > emit

Line 98: expecting > expected

Line 98: principles > principle

Line 204: “In a first hand” > First