Landscapes in the Critical Zone: Towards Geo(Morphic) Large Language Models in the Digital Earth

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript proposes elevating decimal latitude–longitude (dLL)  coordinates from simple geolocation identifiers to “information tokens” and combining Article(+) with dLL to retrieve information from figures and tables. However, the central ideas largely build on existing concepts. In essence, the manuscript mainly advocates the standardized inclusion of geographic coordinates in scientific figures and tables. 

1. The manuscript contains serious formatting and layout problems, particularly on pages 1, 3, 4, 14, and 18. These issues substantially affect readability and should be carefully corrected.

2. The manuscript cites a large body of the authors’ own work published between 2022 and 2026. As a result, it reads more like a summary and promotion of the authors’ previous research than a balanced and objective review of the broader field. 

3. The manuscript lacks empirical validation. No quantitative experiments, comparative evaluations, or case studies are provided to demonstrate the effectiveness of the proposed approach. For example, the authors should quantify the extent to which dLL-based representation improves information retrieval and compare it with established geospatial annotation formats such as GeoJSON and Well-Known Text (WKT). A complete workflow—from data annotation and indexing to querying and retrieval—should also be presented and evaluated.

4. The discussion of how Article(+) is integrated with dLL remains vague and underdeveloped. At present, the manuscript appears merely to suggest using dLL coordinates as row headers, without clearly explaining the underlying mechanism and implementation. 

Author Response

 

The manuscript proposes elevating decimal latitude–longitude (dLL)  coordinates from simple geolocation identifiers to “information tokens”

On a very basic level this is correct and I have used this in several earlier papers to show the idea works in practice. Nobody had done this previously. But [dLL] provide a way of itemising the landscape elements and that, within language models, the [dLL] acts as a geolocated marked for information that needs to be considered as a token.

and combining Article(+) with dLL to retrieve information from figures and tables. However, the central ideas largely build on existing concepts.

Yes, of course on existing concept, but I have gradually extended the ideas in a consistent manner for providing information about digital landscapes in the widest sense. Nobody has done this before. It is perhaps surprising that this had not been done before as reviewers to my ES3 paper of 2024 (for Geol Soc London) mentioned.

In essence, the manuscript mainly advocates the standardized inclusion of geographic coordinates in scientific figures and tables. 

Yes, but even the standardisation is important. I have just finished a review for J Geophysical Research and one for ‘geohazards’ where FAIR data was not applied and there was no standardisation even within a paper for locations to be treated uniformly.

1. The manuscript contains serious formatting and layout problems, particularly on pages 1, 3, 4, 14, and 18. These issues substantially affect readability and should be carefully corrected.

The paper was transferred form another journal and I can only assume that the MS you saw was a bit mangled.

2. The manuscript cites a large body of the authors’ own work published between 2022 and 2026. As a result, it reads more like a summary and promotion of the authors’ previous research than a balanced and objective review of the broader field. 

I accept that there several of my papers referred to but this is because there are no previous uses of the [dLL] method and methodology either for providing unique geolocations, for examining material transfers down potential gradients and for preserving information. Nobody has done this previously.

3. The manuscript lacks empirical validation.

The examples I use, as well as in papers previously, provide empirical evidence and that they ‘work’ in practice is validation.

No quantitative experiments, comparative evaluations, or case studies are provided to demonstrate the effectiveness of the proposed approach.

But the first part of the paper is about comparisons of landforms and gives case studies!

For example, the authors should quantify the extent to which dLL-based representation improves information retrieval and compare it with established geospatial annotation formats such as GeoJSON and Well-Known Text (WKT). A complete workflow—from data annotation and indexing to querying and retrieval—should also be presented and evaluated.

I have already shown in the examples that [dLL] improves information recording and thus FAIR retrieval. It provides an easy way (and that can and should be used in papers and catalogues) to record location. It is a measurement ‘primitive’ (albeit related to WGS24) and is a co-ordinate reference system – which GeoSJON uses! WKT and WKB are a mark up language but still have to manipulate co-ordinates. [dLL] is more efficient and general than W3W and local, map-relate, systems such as ’national grids’, not to mention UTM become non-standard.

 I take your point about supplying a workflow and I have now provided one in summary form as Fig 8 .

4. The discussion of how Article(+) is integrated with dLL remains vague and underdeveloped. At present, the manuscript appears merely to suggest using dLL coordinates as row headers, without clearly explaining the underlying mechanism and implementation. 

 

I have added various bits to improve this, including the new Figure 8

The workflow diagram helps to explain your first point here I hope and I shall supply more on the ‘row header’ aspect.

Reviewer 2 Report

Comments and Suggestions for Authors

The file provided for review contains serious formatting errors, including overlapping paragraphs (pages 1, 3, 4, and 14) and characters superimposed on the text (page 3). Figures overlap the text on pages 10, 16, and 17, and tables are completely misaligned on page 18. These issues made it impossible to read certain parts of the paper.
The paper addresses a highly relevant topic, as it seeks to discuss the incorporation of supplementary elements—linked to the main article via digital identifiers—that can be indexed and accessed through research search platforms and artificial intelligence.
The paper's central focus becomes quite diffuse. By attempting to present a wide range of examples—from agriculture to applications on Mars—the authors dilute the core concept of data tokenization, causing their actual proposal to take a back seat. While I understand the authors sought to justify the need for data tokenization, the sheer diversity of applications distracted from the primary goal: demonstrating the applicability of tokenization and the proposed definitions for these indexing methods.
I would suggest that the authors remove some of these case studies and devote more time and discussion to the proposed indexing elements. The list of suggested information protocols—intended to enhance information mobility—appears only in the paper's conclusions; this is unacceptable. That list needs to be included in the discussion of results, and the analysis should be grounded in it. This list represents the heart of the paper's argument and should not be relegated to the end of the conclusions.
The use of so many different applications makes it somewhat difficult to grasp the proposal and the concept of data tokenization itself. The authors could instead focus on a single example to apply and define their data indexing proposals; this way, the reader would not have to struggle to find connections between disparate examples like glaciers, agriculture, and Mars, but could instead clearly understand and observe a practical demonstration of the proposed approach. Another aspect that could be further developed involves providing more technical explanations regarding the proposal of tuples and the use of brackets—for instance, by demonstrating through practical search examples how algorithms analyze the same case with and without brackets (such as block-based syntax reading versus handling numbers that might otherwise be fragmented or interpreted piecemeal).
The methodology section could be elaborated upon in greater detail. Particularly if all the analyzed examples are retained, a more thorough explanation is needed to clearly present the items analyzed in each situation, demonstrating existing elements and how they would be represented in each application. The methodology itself could also outline the process of conducting a search both with and without these indexing elements.
Regarding the paper's general structure, I recommend a formatting review; most figures overlap the text, and Table 1 is completely misaligned and unintelligible. The conclusions are relevant, though I suggest moving the list of protocols to the discussion section.
The references used are relevant, up-to-date, reflect the current state of the art, and successfully link to the discussion of the results.

 

Author Response

The file provided for review contains serious formatting errors, including overlapping paragraphs (pages 1, 3, 4, and 14) and characters superimposed on the text (page 3). Figures overlap the text on pages 10, 16, and 17, and tables are completely misaligned on page 18. These issues made it impossible to read certain parts of the paper.

Author Response:

Sorry to hear about this. This paper was transferred from another mdpi journal and I used an mdpi template for the layout so I’m not sure what has happened.  However, point noted and I shall ensure that the revised MS will be ok.

The paper addresses a highly relevant topic, as it seeks to discuss the incorporation of supplementary elements—linked to the main article via digital identifiers—that can be indexed and accessed through research search platforms and artificial intelligence.

Thank you for recognising the importance of the topic.

The paper's central focus becomes quite diffuse. By attempting to present a wide range of examples—from agriculture to applications on Mars—the authors dilute the core concept of data tokenization, causing their actual proposal to take a back seat.

Ok, I take your point from a readers pov. However, I added examples to the original (geological-based) MS as I wanted to show, e.g. with respect to a wide variety of subjects and especially for the sharing of spatially-referenced information. We tend to be so wrapped up in specialisms that some general principles of data handling are not regarded.  I have modified the text to allow retention of the examples, and that these all need to reflect FAIR data principles.  The Critical Zone concept (figure. )

While I understand the authors sought to justify the need for data tokenization, the sheer diversity of applications distracted from the primary goal: demonstrating the applicability of tokenization and the proposed definitions for these indexing methods.

My previous comment applies I feel.

I would suggest that the authors remove some of these case studies and devote more time and discussion to the proposed indexing elements. The list of suggested information protocols—intended to enhance information mobility—appears only in the paper's conclusions; this is unacceptable. That list needs to be included in the discussion of results, and the analysis should be grounded in it. This list represents the heart of the paper's argument and should not be relegated to the end of the conclusions.

OK, point taken and I shall reinforce what I have to say about protocols and bring this up front as a separate section as part of the discussion and provide a summary statement in the conclusions. I have also provided, more formalised ‘workflow’ at the end of the discussion to show how the various elements of these ideas come together. This is described more fully in the caption to this new Figure.  

The use of so many different applications makes it somewhat difficult to grasp the proposal and the concept of data tokenization itself. The authors could instead focus on a single example to apply and define their data indexing proposals; this way, the reader would not have to struggle to find connections between disparate examples like glaciers, agriculture, and Mars, but could instead clearly understand and observe a practical demonstration of the proposed approach.

My responses above still apply.  However, I have provided some more explicit statements about linkages. For example, and ref to Figure, the studies of Martian features tend to rely very heavily on terrestrial examples and possible processes involved. Thus, statements need to be quite explicit about features on both planets.  I have also included a brief mention about using directed acyclic graphs, DAGs (currently another topic, dealt with separately from this paper) where the [dLL] geolocation can be used in ‘truth statements’.

Another aspect that could be further developed involves providing more technical explanations regarding the proposal of tuples and the use of brackets—for instance, by demonstrating through practical search examples how algorithms analyze the same case with and without brackets (such as block-based syntax reading versus handling numbers that might otherwise be fragmented or interpreted piecemeal).

Ok, I have added a bit about this. However, a practical example as such is not yet possible as the basic system is only just being implemented.  However, I have provided as an extra paragraph to show how even location identification in the present referencing systems and in ‘inventories’ in particular.  What you suggest is something I am working on and will add but In a later paper.

The methodology section could be elaborated upon in greater detail. Particularly if all the analyzed examples are retained, a more thorough explanation is needed to clearly present the items analyzed in each situation, demonstrating existing elements and how they would be represented in each application.

The methodology itself could also outline the process of conducting a search both with and without these indexing elements.

Fair comment, I have made some more explicit references about this.

Regarding the paper's general structure, I recommend a formatting review; most figures overlap the text, and Table 1 is completely misaligned and unintelligible.

Re my earlier statement. I think you copy of the MS got mangled but I shall check a new version arrived ok with the editors.

The conclusions are relevant, though I suggest moving the list of protocols to the discussion section.

Thank you. Point taken about the protocols as above.

The references used are relevant, up-to-date, reflect the current state of the art, and successfully link to the discussion of the results.

Thank you for this as it was my main purpose to put these ideas before as wide community. Your previous remarks, for which many thanks, will improve the paper.

 

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

Please see the attachment.

Comments for author File: Comments.pdf

Author Response

Thank you for your comments

I have corrected the minor edits

The paper was transferred from another MDPI journal and I was asked to submit to the format given in the MS.  I shall converse with the Editor as to what they would prefer the type now.

I have added various new material to explain aspects better, improve the flow and to provide additional information where necessary.  This is marked in Green on a revision document submitted together with a new Figure 8 and added references to extend the scope of the paper.

Tokenisation.  I fully appreciate that in LLMs map token to embeddings to carry out txt manipulation but (and perhaps I need to emphasise this more) a LLM is not really going to cope with any specialised tokenisation so the geolanguage model requires gentized RAG as well as semantic reasoning to deal with diverse information provided from geological  papers and often vague ontologies – let along combine any of the diffeences in ontologies with other subject.  I have added a bit more to the text on this.

I have modified the text to explain things a bit better. I didn’t want to become too technical but I agree some additional material is required and I hope I have supplied this without over-complicating.  

Structural flow

OK, I have modified this and restructured it to make it smoother and less abrupt. I have put some literature review into the introduction

Minor points 1,2, 3 I have modified and corrected these

 

.

 

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The author has provided a number of highly relevant updates. Key elements that previously posed analytical challenges—involving varied discussions due to the diverse application examples used—have become much more consistent, making the text easier to understand and illustrate.
The subject matter has been better contextualized, and the text restructured to clarify the discussions and proposals. Newly added illustrations have significantly enhanced the clarity of the propositions and the proposed results.

Reviewer 3 Report

Comments and Suggestions for Authors

 Accept in present form