Review of "IWV observations in the Carribian Arc from at network of

ground-based GNSS receivers during EUREC4-A, by Bock, Bosser, Flamant,

Doerflinger, Jansen, Fages, Bony and Schnitt

GENERAL

This is a nice paper on the quality of ZTD and IWV estimates, related to 

both processing speed, in the processing and type of auxiliary observations to

reach to IWV. There's also a small part discussing the relation the relation

between IWV levels and patterns versus the occurence of specific cloud patterns

in the region.

The quality of the presentation is high. In fact I found only one thing that

really ought to be changed: There is a lot on information about how various

approaches to the processing impacts ZTD and IWV quality, but reading the 

title potential readers will not appreciate that.  The remedy is to either

come up with a longer more precise title, or split the material up in two. In the

latter case with a short description of the processing and the discussion 

with respect to clouds patterns, leaving the detailed accounting about quality

versus processing for another manuscript.

SPECIFIC

Early on tell that Grantley Adams airport is on Barbados.

Refer to BCON and BCOS already in section one, in order that all readers 

understand where Barbados is when you refer to fig 1.

Mention the proximity of BCO and the airport with the additional radiosone

early on.

On page 6, move the sentence starting in line 153 up above line 150.

Consider if you later uncertainty assessment requires the text in line 150-153,

and if so, the text should be placed elsewhere.

Line 199: Restrict the comment about ecmwf quality to the location. As you mention

yourself later, ecmwf is grossly off wrt. topography at several of the other GNSS

sites.

Line 253: Would it be healthy to relax the constraint?

page 17 and further. Several places you mention subpages of page 13 and 14,

but they are not indicated on the figures. It should be OK just to refer

to fig 13 and 14, otherwise introduce the a, b, c.. division somehow.

Section 5 appeas to be partly discussion and partly conclusion.

Line 618  Antenna phase center models can result in a global bias

I think. At least we saw a change in biases between GNSS and meteorological 

ZTDs when a significant change was made some years back.

fig 9 Consider adding the location names

fig 14 Enlarge the GNSS circles to make the figure easier to read.

Table 2: It would be good to introduce a blank row before the IWV part of the

table.

This paper gives a very detailed description of GNSS measurements from 49 stations that were obtained during the EUREC4A campaign in the Caribbean Arc. The authors produced a 5 minute-temporal resolution dataset for each station for the purpose of scientific studies. The data for each station are available online in properly formatted user friendly NetCDF files. The authors compare this data to 4 operational streams of GNSS measurements, as well as to collocated measurements from a microwave radiometer and atmospheric soundings (two different types of sondes), as well as to ERA5. To demonstrate an application of the dataset, they link the profiles of integrated water vapor to different cloud patterns that occurred during EUREC4A. The description of all data and the data inter-comparison is done with great care. The authors report very interesting biases between the different data sets and provide hypotheses that may help explain the biases, even though no final answer is given, which is fine. This work should be highly useful to users, especially as the authors make specific suggestions which of the five GNSS streams might be most suitable for certain applications. In only have very minor comments.

Minor comments:

• The authors suggest that biased GAIA observations caused a bias in ERA5 – a very interesting result. Is it possible to find out if they really were assimilated? BCO soundings were also assimilated in various operational analyses, but they do not have a dry bias. Given that BCO soundings occurred more frequently than GAIA soundings, that suggests ERA5 assimilated only GAIA, but not BCO? It would be really nice to know more about this and therefore I would like to encourage the authors to track the path of the data through the assimilation system of ERA5.
• Already in the abstract, Flowers, Fish and Gravel are mentioned. I doubt that all readers are familiar with what they are. Perhaps keep it more general here and just speak of cloud organization.
• In the data there is geoid height and ellipsoid height. Could the authors please define the meaning in the text?
• A4 in https://essd.copernicus.org/articles/13/491/2021/ shows that the air over sea is not more moist than that over Barbados. This is in contrast to the hypothesis at line 355.
• Figures often use a combination of green and red. Please ensure that this is color-blind friendly.

Typos:

Line 82: were -> where

Line 633: biases -> biased

Line 666: spatio-spatial ? & general good -> general in good