Dear Alexander Kmoch and co-workers,

Thank you again for contributing to our special issue. Please accept my apologies for the recent delay in the processing of your manuscript. After re-studying the previous referee reports and the current ones, I come to the conclusion that your really have to clarify your manuscript to work out the valuable results you present.

I have full confidence that you have all the material ready to make this a really good contribution to the field. Generally speaking I would suggest to remain much more specific to the issue of your data and leave out most of the reference to SWAT modelling and pedotransfer approaches. This could not only make your study much more easy to understand but would highlight on the links between soil data, mapping and landscape functioning - as is the topic of our special issue.

Please see the referee reports attached to guide your revisions. In case of question, please do not hesitate to contact me.

All the best.
Kind regards,
Conrad

Dear Alexander and co-workers,

Thank you very much for your intense work on the manuscript. It really gained a lot more conciseness. I am now in the unpleasant situation to ask for another (hopefully largely technical) iteration before I can send it to the reviewers. The reason for this is that I anticipate rather superficial comments since the flow of the manuscript is not yet to the point. I am fully confident that the following suggestions are easy to work through and that it will help the next iteration to finally get the manuscript into a publishable shape. I apologise to cause you work but I am sure it is to the best of your manuscript.

Abstract:
P1L22-24 (I refer the pages and line numbers to the annotated version in your reply letter) is the most important point here (in my view). The first paragraph appears to be more a relict of the first version. I also do not understand in which way the scales are of importance. Later in the manuscript you relate to SoilGrids as reference but this comparison is not at all summarised in the abstract. Moreover, I was wondering if the title might be amended to somehow relate to the large sources in addition to the "Soviet Soil Map".
What I have in mind is especially that the situation you worked through is not uncommon. There are soil maps in many countries but the eco-hydrological process properties/parameters are very difficult to distill. You provide one example how to do this plus the final data to use SWAT in Estonia. So the central aspect of your manuscript should really emphasise on this as an example.

Links as citations:
Although I can understand the notion of using the weblinks in the manuscript, I think the proper way would be to cite the sources similar to paper references. Websites will also require a date for the last access. Please see the guidelines here: https://publications.copernicus.org/for_authors/manuscript_preparation.html

Introduction:
P2L4-6 are central to me. I consider it a better start of the manuscript to state the problem instead of referring to SWAT. This also immediately opens your story: There is detailed soil information but eco-hydrological and pedi-hydrological process modelling (and understanding) requires information on different properties at finer scales. Then you can generalise what information exists in soil maps and global datasets and at what resolution.

I prefer to give the scale triplet (Western and Blöschl, 1999) instead of a map scale which I have little idea about the actual data density and validity. At the moment there are 1:10000, 30 arcsec, 250m, 1:500000, 1km etc. given. But I find it very difficult to get an idea about these resources, how the data has been derived and for which scale it might be applicable. I.e. remote sensing of soil properties in temperate climate is not very reliable nor can it give information about the soil profiles over depth. However they provide the advantage of spatially continuous data. You also point to the many soil sampling campaigns leading to the existing soil map. The soil samples are exactly opposite to remote sensing as they provide very detailed information for some points but it remains challenging to transfer this information to a spatially continuous map. I see this in close relation to the overview about data you use as additional inputs and data which is used as references plus the actual challenge of your study. Moreover, this is a fundamental link to our special issue I really think you should exploit for your argumentation.

The introduction features so many possible data sources but it actually gives no outline of your manuscript. This comes P6L10ff. Somehow also 2.1 is hardly discernible from the introduction this way. I propose to restructure here. The introduction needs to guide the reader into your topic. The long list of soil-related resources can be the first subsection in the methods section. You may consider the many possible resources to be reported as a table? Please also clarify what information will be used to derive your dataset and which is simply reported for referencing and to clarify the importance of your study. Also the PTFs are not clear why they are given extensively in the introduction. I suspect that this can become one subsection in the methods and again detail the ones you really used and give the other references as brief as necessary.

Methods:
As stated above, Figure 1 should come rather early. As you have done, the methods then follow the four steps. And as proposed earlier, I suggest to include the respective background information, references etc. in these subsections. So everything about the actual mapping, unit derivation, etc. is maybe 2.1, 2.2 is A etc.

All Figures should be included as a vector graphics if possible. Fig. 1 still holds v1.0 and other details which might be easy to clarify in a revision. You said USLE_k is omitted as intermediate parameter? I would also suggest to reduce the used symbols to a minimum and to give a legend. The 4 main boxes could simply get their respective headers. The dashed sub-boxes and speech bubbles might simply become the same shape as annotations? The resulting properties could be given as something like a bottom line?

In the methods, the subsections should really align with the workflow in Fig. 1 to avoid confusion. If necessary, details can be of cause extended in subsubsections. Make it easy for the readers to follow. Always trace what information is really used and what is further reference. Please try to give it always the same pattern: What is needed, available sources, other examples, how did you derive the property and how will you evaluate the results. Please avoid jumps between sub-topics. I think it is all there but still a little hard to fit the puzzle. E.g. 2.2 is about texture. Texture can be encoded as fractions of sand, silt and clay (or finer textural classes) or as WRB name. The layers (2.2.3) are for me a different thing more related to the geometric questions of mapping.

2.3 starts off with reference to 2.4. Maybe this could be more content of the overall introduction when Fig. 1 is explained in detail? Then 2.3 is clear to derive topographic indices and you have room to also link to the method background why these indices are selected. Maybe it is not even necessary to use so many subsections?
In 2.4 you again motivate the application of PTFs of 2.5 but the overall pathway could be given earlier to avoid confusion and to really concentrate on SOC, SOM and BD. Since you use a random forest, I would simply remain with RF as one way of machine learning. Avoiding to use the two terms as synonyms could give room for details and clarity. I think that it is important to also clarify how and when PTFs (which are often also ML e.g. Rosetta as ANN) are used. Are they really part of 2.4?
2.5 is very brief although I suspect it to be the very crucial step. Contrary to the former subsections which should be more concise, I think here could be a little more explanation e.g. why the AWC, FC, WP are important. Moreover, again some layers are defined. This time at fixed depths. I still struggle to understand the geometric idea of your database.

I think the final database including the respective names and encodings could be a further section after the methods. There can come all details about your naming conventions etc.

Results:
I understand the results to align best along the four subsections again. At the moment I really struggle to follow how the evaluation is done and if the maps are somehow reliable. Maybe you could use a few of the sampling areas from Fig. 2 plus a few other areas with information from other resources here? The maps of course look beautiful. But I have no idea if they achieve your goal for SWAT modelling.

Figure 9 is clearly result evaluation. But you refer to it only in the methods 2.2.4. I like the plot but cannot read the axes. I think this plot is too small. Maybe you can think of another form to give this reference? Moreover, I am a little puzzled about the continuous histograms in SoilGrids but the very few distinct bars in your database for texture. Also the other parameters do not really appear to resemble SoilGrids. I am not at all saying that SoilGrids is more correct. But you really have to discuss why you come up with different values and why your data is trustworthy.

Discussion:
Maybe you could move some of the open issues from the methods section to the discussion? I would expect a little more structure what is discussed here.

Again, I am sorry to return the manuscript before sending it to reviewers. I simply fear that we might enter an internal loop of rather structural deficits which obscure to dive to the actual matter of your work. I hope you can easily follow my suggestions and that you find it helpful. In case of any questions, please contact me. We can even quickly arrange a video-call.

All the best.
Conrad

Dear Alex and co-workers,

Thank you for all the efforts you have invested into the manuscript. I have received the reviews of one anonymous referee and went through the manuscript myself. There are a couple of minor issues left but this should be really easy to resolve. I hope you can read all the comments of the first referees and inside the pdf. If there are any difficulties, please let me know.

I am very happy to receive you revisions soon so the manuscript can be finally accepted.
All the best.
Conrad

Dear Alex and co-workers,
Thank you for the tremendous work you have invested. I think it was worth the effort. I am very happy to accept your contribution to our special issue.

I just spotted a couple of minor technical issues, I ask you to resolve before handing the MS to the copy editing and typesetting.

All the best and kind regards.
Conrad


P3L20: RF is used for the first time here. I suggest to simply drop it. I still find the list of topics in the paper difficult to comprehend due to the shift in sentence structure. Moreover this paragraph sets off to give the structure of the paper but only drafts the next section 2. Pls. check.

P3L25: There appears to be a reference error of Latex.

P14L11: the last number is with comma as thousand separator. I do not know what is the house standard at copernicus but I think it is given in the author guidelines. Pls. check.

Fig 5: There are comma instead of dots as decimal separators...

P21L18: improvement > improved?