1. There are quite  a lot of grammatical and style issues with the presentation that distract from ease of reading. There are also many examples of non-scientific terms such as 'pretty' that are going to confuse people for whom English is a second language.

2. The rationale for the tool is good, and is well explained.

3. Paragraph beginning line 69: several statements are made that read more as opinions instead of evidence. Given that this is where you are justifying a large part of the effort required for involvement, I think your statements need to have good evidence here.

4. There are probably as many different ways to design this kind of system as there are soil scientists. Please give a bit more justification for the structure of the overall system, particularly in terms of which factors are considered important for data entry.

5. The Technical Implementation section needs to come much earlier, so that the reader can visualise not just what you have done, but how you have done it. I suggest having it as section 3.2.

6. The Technical Implementation section does not contain enough detail for the work to be duplicated.

Thank you for letting me review this manuscript. I would like to congratulate the Authors on the good work put into this manuscript and overall on the KLIB project!

The article is well written and the structure is adapted to present a tool (the KLIB) that is certainly useful for the soil science community. I myself used this tool in several projects and I am happy to see it further developed! I agree with the authors that often the outcomes of a specific agricultural practice are bound to specific pedo-climatic conditions. To my knowledge, the KLIB is one of the few (if not the only one) tools that enable us to access this pedo-climatic information in a structured way. The authors of the KLIB propose, based on publications, to extract this information (manually for now, more automatically in the future) from papers and make it available in different visualization tools based on network graphs. This is certainly of interest at times where the scientific literature is growing quickly. I hope this publication will give it more visibility.

Below are a few comments and suggestions:

L7: " a number of visualization tools": maybe be more specific with some examples?

Abstract: I would directly add the URL of the KLIB so interested person can have a look at it directly from the abstract

L47: 'correlation' is repeated twice in the sentence

L71 double negation: "not-inconsiderable", maybe "not negligible" or "considerable"

L80: I agree, we need a way to report information with structure and the KLIB is a good option for that (among other things)

L102: for organic carbon: is it possible to also specify the method used (and possibly the date of the sampling)? values can varies quite a lot between methods

L105: I would give a few examples of a drivers and measured variables already here in parenthesis

L110: lovely! I like standard terms!

L110: does the tree enable intermediary terms selection? For instance if I know pH was measured but there is no information in the paper if it was with the pH KCl or pH H2O method, can I still add it as a more global level? like just pH? And in this case, if I select just ‘pH’ in the keyword tree, will I be able to see the publications for both ‘pH KCl’ and ‘pH H2O’? I miss the information about how it works with the ‘levels’ of the keywords tree throughout the KLIB

fig1 is not referenced in the text

fig2c I always find it strange that the property and drivers have the same color (of the dot). Wouldn't it make sense to separate drivers and measured variables? or is that intentional?

fig2 and fig1: add a date when the screenshot was taken (as it might evolved in the future I guess) .. for next figures too

PC mode and others mode: there is always a "share at least 2 properties" rule, can this be modified (to 3 or 4 properties for instance?) - this might be a future feature request

L175: why not also add a .csv output? (easier for people working in spreadsheet) to be able to download the entire list and not just entry per entry (a JSON/CSV combination can be nice)?

L185: this is a good example. But it supposes that the user knows (or tries a few times) to see on which pedo-climatic variables the relationships vary. Isn't a way to show (a bit similar to a meta-analysis moderator figure but in a qualitative way) the direction of a relationship according to different pedo-climatic variables. We could imagine we select 1 driver (and a reference possibly), 1 soil variable and then several pedo-climatic variables and we make a figure of all that? 

L210: "it is important not to mix ... experimental manipulation": this sentence and how it articulates with the previous one is a bit confusing for me. Are the "relationships identified through experimental manipulation" linked with "theoretic hypothesis (focus on causality)", if so I suggest switching the order of one or the other sentence so that it is consistent between sentences

L217: I agree that KLIB can help access this information but I still find it difficult to have an overview of the pedo-climatic factors and their influence on a given soil-property vs practice relationships. KLIB enable filtering and then we can explore each list of publication individually but we don't have an 'overview' (similar to a moderator figure you can find in meta-analysis when an effect size is decomposed per soil group for instance) see also L185

L229: "screen by pairs of the scientific community" what is meant by this? only pair-review publications are accepted or is there another kind of filtering behind? according to which criteria? Is there a double quality control on new entries inserted?

section 3.5: nice to have the implementation description. I was wondering why you went for a mongoDB (unstructured database) rather than a (traditional) relational one? Also what kind of consistency control does mongoDB implement? (relational have "ACID checks" -- I am not an expert on this)

L295: I was wondering if it wouldn't also be interesting to encourage manuscript's author to add their publications to the KLIB? Another proposition could be to enable a “second check” by another user upon entering data in the KLIB to ensure quality control (but I understand it takes time). Working with the transparency of who added the entry is already a good start as you mentioned.

L301: I would also add the fact that a cookbook page is also available on the website and serve as documentation for adding new entries

L331: Angular reference, a bit strange to cite it like this but why not. Then other frameworks should be cited too I guess no?

Open questions:

The database collected in this work is certainly of interest and can be used for further research. I think it can be good to add a way to download this database and list of selected papers. Being able to explore it through the different visualizations is already very nice but does not allow for more automated methods to be applied.

Does the KLIB have a DOI itself? Or shall we cite this paper if we want to use it? I suggest adding a “how to cite” section to the website maybe in the ‘info’ tab. This can be added to a ‘data availability’ section of this manuscript.

Is the KLIB only based on field experiments?

How about review papers or meta-analyses? Can they be entered in the database or is it mainly “primary studies” that are seeked?

If a publication contains two experiments made on different soil types (so two sites) and that different relationships have been found for each type, should it be entered twice so that the two different relationships can be added?

Bug: sometimes the link to Google Scholar is not populated. How to reproduce: in ‘Properties affected by management’, type ‘bulk density’, then select one of the gray dots and click on the “external link icon” for Google Scholar.