Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Review v1 on the manuscript:

Iterative Score Propagation Algorithm (ISPA): A GNN-Inspired Framework for Multi-Criteria Route Design with Engineering Applications

This manuscript deal with the problem of deciding rote trajectories. Determining near optimal trajectories of routes is a classical problem. With the advent of computational power, we may attempt to solve these multidimensional decision problems as optimization problems. However, computing power does not solve everything. This study analyses several methods to implement optimization problems highly complex.  Attention is given to the criteria used to establish the objective function. The study proposes a GNN-inspired algorithm named Iterative Score Propagation Algorithm (ISPA). Comparison with other decision making techniques as AHP, TOPSIS, VIKOR, and entropy assessments are performed. The results of the tests implemented show the ISPA performs better. Besides this improvement, the authors highlight the method proposed offers the advantage of being a spatially-aware procedure.

General comments

I myself identify with the authors intention to pursue “deterministic and interpretable models”. I think it is a very appropriate phrase to end the Introduction Section.

The study evaluates three construction projects dealt with two groups of techniques: the first group considered objective, and the second group tagged as subjective. In the group of objective techniques, the proposed algorithm (Iterative Score Propagation Algorithm, ISPA) is included along other techniques as the classical Weighted Linear Combination (WLC), TOPSIS, and with VIKOR. In the group of subjective techniques the study includes AHP as a procedure to produce the weights that will govern the solutions of all different approaches tested. Thus, I do not see parallelism in the results shown in section 4.4 with respect to the results 4.3. In other words, I do not see why AHP are shown together with  ISPA, TOPSIS and VIKOR results, and why are WLC appears as another technique., If WLC and AHP are just two different ways to produce weights, then they should not be included at the same level as ISPA, VIKOR and TOPSIS. Maybe I am just confuse or do not understand the structure of the set of experiments. I suggest including a diagram with this structure explaining all the problem approaches applied to these three construction projects. I think it might be an diagram with the shape of a tree.

Another presentation aspect I think is fixable is the description  of the ISPA, The manuscript does not show the ISPA. Being the ISPA the central element of this study, I regard not acceptable missing details of how it works. Section 3.2.3 describes some aspects of ISPA. I suggest adding the corresponding pseudo-code or a figure showing the major blocks of the algorithm.

About form and writing

In section 5.2 the authors argument that traditional methods are rigidly bound to cost minimization. While this statement is the authors opinion, and of course that is valid, this sentence may be regarded as a subjective statement, that might be considered speculation, or at least, a non-proven fact. My understanding is that the optimization paradigms rely on the ability of analysts to identify the parameters affecting the objective function and to describe (or concentrate) in one number the resulting compromises existing in a complex optimization situation.

The last paragraph of the document: “In conclusion, this work offers more than just a novel algorithm; it proposes a more flexible, transparent, and spatially intelligent approach to route optimization. The ISPA framework holds significant potential as a powerful decision-support tool for planners and engineers, enabling the design of more efficient, more sustainable, and even more experiential infrastructure corridors that better balance the priorities of diverse stakeholders” does not belong to a scientific document. None of the attributes mentioned in this paragraph were quantified. Moreover, this problem is present in all conclusions paragraphs. I must respectfully say using this style of writing does not show respect for the reader and for science. I believe this is not intentional, however the quality of the manuscript is severely affected, in my opinion.

Whatever the source of this rhetoric style of writing is, I suggest the authors pruning the document and limit to present only the objective unadorned facts they can report.

Final comments and decision

This manuscript proposes an interesting decision tool to handle complex optimization problems. The experiments to assess the suitability of this new tool is complex because the experiment and the systems used for this evaluation are complex themselves. The manuscript already offers a good structure to depict the study goals and results. However, I think there is a good opportunity to make the manuscript more effective. After some presentations improvements, this manuscript may deserve publication

Author Response

Comments 1: [The study evaluates three construction projects dealt with two groups of techniques... Thus, I do not see parallelism in the results shown in section 4.4 with respect to the results 4.3. In other words, I do not see why AHP are shown together with ISPA, TOPSIS and VIKOR results, and why are WLC appears as another technique. [...] I suggest including a diagram with this structure explaining all the problem approaches applied to these three construction projects. I think it might be an diagram with the shape of a tree.]

Response 1: Thank you for pointing out this critical ambiguity in the methodological structure. I agree that the distinction between Weighting Philosophy (AHP/Entropy) and Scoring Method (WLC/TOPSIS/VIKOR/ISPA) was not clear, particularly regarding the dual role of WLC. I have implemented the following changes to ensure transparency:

1. Diagram Revision (Figure 1): Figure 1 has been entirely revised to visually articulate the dual-stream (Objective vs. Subjective) parallel analysis. The revised figure clearly shows Criterion Weights (AHP or Entropy) feeding into the Parallel Suitability Scoring box containing the four aggregation methods (WLC, TOPSIS, VIKOR, ISPA).
2. Method Name Standardization (Figure 1): We standardized the method names in Figure 1 to use WLC instead of AHP for the weighted summation method, clarifying that WLC is the calculation engine used under both weighting philosophies.
3. WLC Role Clarification (Text): We explicitly clarified the dual capacity of WLC in the text:

[Revised Text in Manuscript]:

Line	Original Text	Updated Text
331–335	This method operates on a fully compensatory strategy... WLC serves a dual role: (1) when combined with objective Entropy weights, it acts as a baseline model... and (2) when combined with subjective AHP weights, it constitutes the suitability mapping step of the widely known AHP-LCP approach.	This method operates on a fully compensatory strategy... WLC is used in a dual capacity: (1) when combined with objective Entropy weights, it acts as a baseline model... and (2) when combined with subjective AHP weights, it serves as the core suitability mapping component of the widely known AHP-LCP approach.

Comments 2: [Another presentation aspect I think is fixable is the description of the ISPA. The manuscript does not show the ISPA. Being the ISPA the central element of this study, I regard not acceptable missing details of how it works. [...] I suggest adding the corresponding pseudo-code or a figure showing the major blocks of the algorithm.]

Response 2: I agree completely. The interpretability ("white-box" nature) of ISPA is a key contribution, and its mechanism must be fully transparent. I have introduced a formal pseudo-code block directly into Section 3.2.3 (Iterative Score Propagation).

[Revised Text in Manuscript]:

Line	Original Text	Updated Text
392 (New Block)	[...] allows information to propagate over longer distances. This iterative process effectively acts as a low-pass spatial filter [35]...	[...] allows information to propagate over longer distances. Algorithm 1: Iterative Score Propagation Algorithm (ISPA) Input:   G = (V, E): Graph Network   H^(0): Initial suitability scores (WLC output)   K: Number of iterations   alpha: Smoothing factor (Propagation factor, set to 0.5)   d_max: Neighborhood search distance Output:   H^(K): Final spatially-aware suitability scores 1: Initialization: Set k = 0 2: Repeat for k = 0 to K-1: 3:   For each node i in V: 4:     Identify the neighbor set N_i based on d_max. 5:     Calculate the mean score of neighbors:            M_i^(k) = (1 / |N_i|) * SUM_{j in N_i} h_j^(k) 6:     Update the node's score using the propagation rule:            h_i^(k+1) = (1 - alpha) * h_i^(k) + alpha * M_i^(k) 7:   End For 8:   Set k = k + 1 9: End Repeat 10: Return H^(K) … This iterative process effectively acts as a low-pass spatial filter [35]...

About Form and Writing

Comments 3: [In section 5.2 the authors argument that traditional methods are rigidly bound to cost minimization. While this statement is the authors opinion, and of course that is valid, this sentence may be regarded as a subjective statement, that might be considered speculation, or at least, a non-proven fact.]

Response 3: I accept this critique and agree that the original phrasing was overly subjective. I have revised the statement in Section 5.2 to maintain a neutral, evidence-based tone, focusing on the structural inclination rather than absolute rigidity.

[Revised Text in Manuscript]:

Line	Original Text	Updated Text
814–815	Traditional methods are rigidly bound to a "cost minimization" paradigm.	Traditional methods are often structurally inclined towards a "cost minimization" paradigm.

Comments 4: [The last paragraph of the document: “In conclusion, this work offers more than just a novel algorithm; it proposes a more flexible, transparent, and spatially intelligent approach to route optimization. [...] None of the attributes mentioned in this paragraph were quantified. [...] I suggest the authors pruning the document and limit to present only the objective unadorned facts they can report.]

Response 4: I sincerely thank the reviewer for highlighting the use of rhetorical language, which severely affected the quality of the manuscript. I have rigorously pruned the entire document to eliminate unquantified, promotional language, focusing strictly on objective, evidence-based claims.

1. Conclusion Pruning: The criticized final paragraph (Lines 874-879) has been removed.
2. Abstract Revision: The Abstract has been rewritten to replace rhetoric with quantified results (e.g., highest overall mean performance (0.629) and greatest stability (1.000)).

[Revised Text in Manuscript]:

Line	Original Text (Example)	Updated Text (Example)
874–879	In conclusion, this work offers more than just a novel algorithm; it proposes a more flexible, transparent, and spatially intelligent approach to route optimization... (Full paragraph)	(The entire paragraph has been removed.)
81–91	Accordingly, the primary contributions of this work are threefold: Methodological Innovation: We introduce a hybrid methodology... An Interpretable "White-Box" Model: We deliberately depart...	Accordingly, the primary contributions of this work are threefold: Methodological Innovation: The study introduces a hybrid methodology... An Interpretable "White-Box" Model: The framework deliberately departs...

Comments 5: [Lines 81-91: These sentences begin with "we," but the paper lists a single author. Please clarify the use of plural pronouns.]

Response 5: I apologize for the inconsistent use of plural pronouns. I have reviewed the entire manuscript and corrected all instances of "we," "our," and "authors" to use the singular pronoun "I" or appropriate neutral phrasing ("this study," "the framework") to reflect the single authorship.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript is a highly professional study on Multi-Criteria Route Design with strong practical relevance. It presents rigorous arguments, and the conclusions are well supported by the data. There are some comments:

 

1. Some expressions in the paper could be standardized or reformatted, such as CSlope (line 190), Stdperf (lines 487, 489), Sstability (lines 487, 491), Sfinal (lines 496, 498), Pavg (lines 483, 484), etc.

 

2. In Equation (2), the authors mention a buffer threshold (δroad = 50 m). Why was 50 m chosen instead of another value, such as 100 m? It would be helpful to provide further explanation or justification for this choice.

 

3. In lines 210–212, the authors state: “This function converts the raw Euclidean distance (𝐷(𝑝𝑖,𝐻𝑛𝑒𝑎𝑟𝑒𝑠𝑡)) of a node pi from the nearest settlement (Ynearest) or sensitive area (Hnearest) into a standardized proximity cost score in the range [0, 1].” The phrase “nearest settlement (Ynearest) or” may cause ambiguity. It is recommended to remove or rephrase this part for clarity.

 

4. Regarding Equation (14), uphill gradients increase fuel consumption; however, when traveling in the opposite direction, these slopes become downhill. This operational penalty may require further clarification or discussion.

 

5. In Table 2, the criterion Croad (Proximity to Roads) changes under different scenarios. The authors could provide additional explanation about the rationale behind these adjustments.

Author Response

Comments 1: [Some expressions in the paper could be standardized or reformatted, such as CSlope (line 190), Stdperf (lines 487, 489), Sstability (lines 487, 491), Sfinal (lines 496, 498), Pavg (lines 483, 484), etc.]

Response 1: Thank you for this essential suggestion. I agree that consistency in notation is critical for clarity. I have reviewed the entire manuscript and standardized all metric and variable acronyms to adhere to formal mathematical notation (using subscripts).

[Updated text in the manuscript]:

•          (Line 190)
•           (Lines 483, 484)
•          (Standard deviation, replacing     ) (Lines 487, 489)
•          (Lines 487, 491)
•          (Lines 496, 498)

Comments 2: [In Equation (2), the authors mention a buffer threshold (  ). Why was 50 m chosen instead of another value, such as 100 m? It would be helpful to provide further explanation or justification for this choice.]

Response 2: I appreciate the request for detailed parameter justification. I have added an explicit explanation for the choice of the 50 m buffer threshold, relating it to common engineering practices and the input data resolution.

[Updated text in the manuscript]:

Line	Context	Updated Text
200–201	This distance was then compared against a buffer threshold ( ) to assign a binary parameter...	This distance was then compared against a buffer threshold ( ), which was selected as it reflects a typical right-of-way width for rural highway projects and ensures local detail is captured relative to the 30m DEM resolution, to assign a binary parameter...

Comments 3: [In lines 210–212, the authors state: “This function converts the raw Euclidean distance (  ) of a node pi from the nearest settlement ( ) or sensitive area (  ) into a standardized proximity cost score in the range [0, 1].” The phrase “nearest settlement (  ) or” may cause ambiguity. It is recommended to remove or rephrase this part for clarity.]

Response 3: I agree that the dual use of the term in the description of the distance function caused ambiguity. I have clarified the text by introducing a general term,    (Feature of Interest), which encompasses both specific criteria.

[Updated text in the manuscript]:

Line	Context	Updated Text
210–212	...distance (  ) of a node   from the nearest settlement (  ) or sensitive area (  ) into a standardized proximity cost score...	...distance ( ) of a node  from the nearest feature of interest (  ) (where   is  for settlements or for sensitive areas) into a standardized proximity cost score...

Comments 4: [Regarding Equation (14), uphill gradients increase fuel consumption; however, when traveling in the opposite direction, these slopes become downhill. This operational penalty may require further clarification or discussion.]

Response 4: Thank you for requesting further clarification on the direction-dependent cost. I have explicitly reinforced the discussion of the

       

     

term in Section 3.3.2 to highlight its role in modeling asymmetric travel cost/reward.

[Updated text in the manuscript]:

Line	Context	Updated Text
443–446	A positive value for               signifies a penalty (e.g., increased fuel consumption), while a negative value can represent a reward (e.g., the satisfaction of reaching a scenic peak on a trekking trail). This parameter is critical to the framework's flexibility and its ability to adapt to different scenarios.	A positive value for               signifies a penalty (e.g., increased fuel consumption), while a negative value can represent a reward (e.g., the satisfaction of reaching a scenic peak on a trekking trail). This parameter is critical to the framework's flexibility and its ability to adapt to different scenarios.

 

Comments 5: [In Table 2, the criterion    (Proximity to Roads) changes under different scenarios. The authors could provide additional explanation about the rationale behind these adjustments.]

Response 5: I agree that the scenario-specific rationale for   (Proximity to Roads) should be clearer. We have added explanatory text immediately following the description of the AHP weights for the Pipeline scenario to clarify why this criterion is defined as a Cost (C) in this specific context, despite its typical definition as a Benefit (B) in route planning.

[Updated text in the manuscript]:

Line	Context	Updated Text
577–578	In the Pipeline scenario, Proximity to Roads, which has no operational significance, is almost disregarded (5%), while priority shifts to the criteria representing geotechnical risk (Slope and Proximity to Sensitive Areas).	In the Pipeline scenario, Proximity to Roads, which has no operational significance, is almost disregarded (5%), while priority shifts to the criteria representing geotechnical risk (Slope and Proximity to Sensitive Areas). This prioritization reflects the consideration that for pipelines, proximity to existing road rights-of-way often increases complexity and acquisition costs, thus requiring   to be treated as a Cost.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Abstract

Expand to include all essential components: research context or program, approach and methods, key results, and overall contribution.

1. Introduction

Lines 69–71: Clearly articulate the specific research gaps this study addresses. Identify and discuss the limitations of existing approaches, supported by proper citations.

Lines 74–77: The proposed “new methodology” appears to be more of a contribution than a model. Based on the text, Reference 16 seems to be a conference presentation—does this article use the same approach or methodology? If so, clarification is needed on whether the author is the originator of the proposed methods.

Lines 81–91: These sentences begin with “we,” but the paper lists a single author. Please clarify the use of plural pronouns.

2. Related work

The section references LCP, MCDM, and GNN. Does this reflect an evolutionary sequence, or are these three distinct approaches within route optimization frameworks?

Additionally, explain how “spatial blindness” occurs in previous methods.

Conclude this section with a synthesis of key findings and a clear identification of remaining research gaps. Particularly, connect those gaps to the current study’s methodology and contribution.

It is also difficult to locate supporting discussion or prior applications of WLC, TOPSIS, or VIKOR (which appear 3. Methodology and 4. Results) in the related work section—please expand on how these methods were selected, how the experiment was conducted, and what outcomes were expected.

3. Methodology

3.1.2 Criteria Definition and Rationale

The evaluation criteria appear to be drawn from a study focused on “suitability” analysis (Reference 24). This provides some foundation but may not fully align with route optimization frameworks. Consider strengthening the justification with literature specifically related to route optimization.

4. Results

Ensure all figures are of sufficient resolution and include clear annotations; several are currently difficult to read.

5. Discussion

The discussion section should move beyond restating results and instead provide a thoughtful interpretation of their meaning and implications. It would be helpful to explicitly address whether the study’s objectives were achieved and how effectively the adopted methodology supported those goals. The author could also reflect on the challenges encountered during the process and the limitations of the chosen approach, situating these within the broader field of route optimization.

In addition, the discussion should compare the study’s findings with previous research—indicating where they align, diverge, or extend existing knowledge—and reference relevant academic work to support these comparisons. Ultimately, this section should clarify how the results contribute to current understanding, what innovations they offer, and what gaps remain for future research.

Author Response

Comments 1: [Abstract: Expand to include all essential components: research context or program, approach and methods, key results, and overall contribution.]

Response 1: I agree that the Abstract needed to be more comprehensive and data-driven. I have completely revised the Abstract to include the full research context, the transparent nature of ISPA, and the key quantified results (highest Overall Mean Performance (0.629) and perfect Overall Stability Score (1.000)), while removing rhetorical language.

[Updated text in the manuscript]: Lines 9–29 (Abstract): The abstract was fully rewritten to focus on the methodological gap, the ISPA framework, and the superior, quantified performance across all scenarios.

Comments 2: [1. Introduction, Lines 69–71: Clearly articulate the specific research gaps this study addresses. Identify and discuss the limitations of existing approaches, supported by proper citations.]

Response 2: Thank you for urging me to articulate the research gaps more explicitly. I agree that clearly defining the limitations of existing work is vital to frame the study's contribution. The Introduction now clearly sets up two primary, distinct research gaps that the ISPA framework addresses:

1. Structural Spatial Blindness in MCDM: I emphasize that methods like TOPSIS and VIKOR fundamentally violate Tobler’s Law (Line 51–54), leading to suboptimal route fragmentation.
2. De Novo Synthesis Gap in GNNs: I clarify that while GNNs are powerful for analyzing existing networks, they are rarely applied to the synthesis problem—designing a new network from scratch (Line 61–65).

This two-pronged approach ensures that the ISPA framework is justified not just by a single limitation, but by a hybrid necessity.

[Updated text in the manuscript]:

• Lines 51–54 (Sec. 2.2): Strengthening of the "spatial blindness" discussion.
• Lines 61–65 (Sec. 2.3): Clear articulation of the "de novo synthesis" gap in GNN literature.

Line	Context	Updated Text
51–54	However, despite their utility, these traditional MCDM methods share a fundamental limitation: an inherent "spatial blindness." They evaluate each potential location as an independent entity...	However, despite their utility, these traditional MCDM methods share a fundamental limitation: an inherent "spatial blindness." This limitation is structural: They evaluate each potential location as a discrete, independent entity, largely ignoring the context-sensitive spatial relationships and dependencies described in Tobler's First Law of Geography [7,8]. This often leads to the generation of fragmented and physically suboptimal routes.
61–65	Despite these impressive achievements, a critical gap persists in the GNN literature... The synthesis problem—where an optimal route must be designed de novo... remains a relatively unexplored frontier for GNNs.	Despite these impressive achievements, a critical gap persists in the GNN literature... The synthesis problem—where an optimal route must be designed de novo... remains a relatively unexplored frontier for GNNs. This gap highlights a need for transparent, non-learning-based GNN frameworks for generative spatial optimization.

 

Comments 3: [Lines 74–77: The proposed “new methodology” appears to be more of a contribution than a model. Based on the text, Reference 16 seems to be a conference presentation—does this article use the same approach or methodology? If so, clarification is needed on whether the author is the originator of the proposed methods. Lines 81–91: These sentences begin with “we,” but the paper lists a single author. Please clarify the use of plural pronouns.]

Response 3: I thank the reviewer for addressing these crucial points regarding clarity and academic tone. I apologize for the inconsistent use of plural pronouns, which has been systematically corrected throughout the entire manuscript. I also agree that the distinction between the method's definition and its originality needed clarification.

The following revisions address both issues:

1. Model vs. Contribution: I have clarified that ISPA is a hybrid computational framework implemented as a deterministic optimization engine (model), which is the primary methodological contribution.
2. Pronoun Correction: All instances of "we," "our," and "authors" have been replaced with the singular pronoun "I" or appropriate neutral phrasing ("The study," "The framework") to align with single authorship.

[Updated text in the manuscript]:

Line	Original Text	Updated Text
73–74	This study directly targets this gap by proposing a novel, hybrid computational framework that reimagines the GNN paradigm...	This study directly targets this gap by proposing a novel, hybrid computational framework that reimagines the GNN paradigm not as a black-box learning tool, but as a transparent and deterministic optimization engine.
81–83	Accordingly, the primary contributions of this work are threefold: We introduce a hybrid methodology that integrates the strengths...	Accordingly, the primary contributions of this work are threefold: The study introduces a hybrid methodology that integrates the strengths...
85–87	An Interpretable "White-Box" Model: We deliberately depart from the opaque nature of deep learning to propose a deterministic and interpretable model...	An Interpretable "White-Box" Model: The framework deliberately departs from the opaque nature of deep learning to propose a deterministic and interpretable model...
88–91	A New Application Domain for GNNs: We expand the GNN paradigm from the analysis of existing networks...	A New Application Domain for GNNs: The work expands the GNN paradigm from the analysis of existing networks...

Comments 4: [2. Related work: The section references LCP, MCDM, and GNN. Does this reflect an evolutionary sequence, or are these three distinct approaches within route optimization frameworks? Additionally, explain how “spatial blindness” occurs in previous methods. Conclude this section with a synthesis of key findings and a clear identification of remaining research gaps. It is also difficult to locate supporting discussion or prior applications of WLC, TOPSIS, or VIKOR in the related work section.]

Response 4: I appreciate the reviewer’s request for a stronger, more synthesized theoretical foundation. While these elements were present in the initial submission, I recognize that they were not sufficiently emphasized to demonstrate the coherence of the methodology. I have clarified the structure, mechanism, and conclusion, ensuring the interrelation of the three paradigms is unmistakable:

1. Structure and Selection Rationale: I clarified at the beginning of Section 2 that LCP, MCDM, and GNN are the three fundamental methodological pillars that converge in this hybrid framework (Lines 92–96). I emphasized the role of WLC, TOPSIS, and VIKOR as standard benchmark tools for generating the LCP cost surface (Lines 109–112).
2. Spatial Blindness Mechanism: I ensured the explicit mechanism of "spatial blindness"—violating the principle of spatial autocorrelation—is clearly highlighted (Lines 114–120).
3. Synthesis and Gaps: The final paragraph of Section 2 (Lines 135–137), which synthesizes the three approaches and defines the ultimate goal of the hybrid methodology, has been structurally highlighted and slightly revised to function as a clear conclusion for the section.

[Updated text in the manuscript]:

• Lines 92–96, 114–120, and 135–137: These sections have been structurally marked and slightly edited for stronger coherence (e.g., ensuring pronoun consistency) but contain the same core information, now presented with greater emphasis.

 

Line	Original Text	Updated Text
92–96	(Beginning of Sec. 2)	This section reviews the three fundamental methodological pillars upon which my proposed hybrid framework is built: (1) GIS-based least-cost path analysis, (2) the integration of Multi-Criteria Decision-Making (MCDM) methods and their inherent spatial limitations, and (3) the rise of Graph Neural Networks (GNNs) as a spatially-aware paradigm.
114–120	However, despite operating on spatial data, these traditional MCDM methods are inherently spatially blind. They evaluate each pixel or node as a discrete, independent entity. Their decision logic operates on the attribute values of the pixels but disregards their spatial or topological relationships. This violates the fundamental principle of spatial autocorrelation [7], leading to a critical flaw: the generation of fragmented and suboptimal routes...	(The original, detailed explanation of the mechanism for spatial blindness remains, confirming the mechanism is explicit.)
135–137	(End of Sec. 2.3)	Building upon these theoretical foundations, the next section details my hybrid methodology, which aims to integrate the strengths and mitigate the weaknesses of these three paradigms (LCP, MCDM, and GNNs).

Comments 5: [It is also difficult to locate supporting discussion or prior applications of WLC, TOPSIS, or VIKOR (which appear 3. Methodology and 4. Results) in the related work section—please expand on how these methods were selected, how the experiment was conducted, and what outcomes were expected.]

Response 5: I agree that the rationale for including WLC, TOPSIS, and VIKOR should be established in the Related Work section (Section 2) to clearly frame them as essential benchmarks. The core motivation is two-fold:

1. Standard Benchmarks: These methods are the most widely published and adopted MCDM approaches for generating LCP cost surfaces in GIS/route optimization literature (Lines 48–50).
2. Structural Comparison: They represent the "spatial blind" approach, which is the exact structural limitation ISPA aims to overcome. Comparing ISPA against these established point-wise scoring methods (and WLC in the AHP-LCP standard) allows for a direct validation of the added value of spatial awareness.

I have updated Section 2.2 to explicitly articulate this rationale, citing the established nature of these methods in the field.

[Updated text in the manuscript]:

Line	Original Text	Updated Text (Lines 109–112, Sec. 2.2)
109–112	In response to this need, MCDM methods such as AHP [18], TOPSIS [19], and VIKOR [20] have become standard tools in the field [3]. These methods generate the composite cost surface required for LCP analysis by combining different criteria layers with weights that reflect decision-maker priorities.	In response to this need, MCDM methods such as AHP [18], TOPSIS [19], and VIKOR [20] have become established and standard benchmark tools in the field [3]. Their selection in this study is based on their prevalence in the literature as the primary means to generate the composite cost surface required for LCP analysis by combining different criteria layers with weights that reflect decision-maker priorities.

 

Comments 6: [3.1.2 Criteria Definition and Rationale: The evaluation criteria appear to be drawn from a study focused on “suitability” analysis (Reference 24). [...] Consider strengthening the justification with literature specifically related to route optimization.]

Response 6: I agree that the justification for the evaluation criteria must be strongly linked to the specific context of route optimization rather than general suitability analysis. While Reference [24] provides a foundation for MCDM criteria definition, I have explicitly reinforced that the four criteria were chosen to rigorously represent the universally accepted four key domains of large-scale infrastructure planning:

1. Engineering / Cost(Slope)
2. Economic(Proximity to Roads)
3. Social(Proximity to Settlements)
4. Environmental(Proximity to Sensitive Areas)

I have strengthened the introductory text in Section 3.1.2 by specifically listing these four core domains and providing additional relevant citations ([2], [25], [26], [27]) that confirm the relevance of these factors in highway and pipeline route design.

[Updated text in the manuscript]:

• Lines 177–182 (Sec. 3.1.2): The text has been updated to explicitly name the four domains and link them to the criteria, clarifying the route optimization context.

Line	Orijinal Text	Updated Text
177–179	Particularly in large-scale infrastructure projects, route planning necessitates an MCDM approach that integrates multi-dimensional factors such as engineering constraints, economic costs, environmental sensitivities, and social impacts [24].	Particularly in large-scale infrastructure projects, route planning necessitates an MCDM approach that integrates multi-dimensional factors such as engineering constraints, economic costs, environmental sensitivities, and social impacts [2,24,25,26,27].
180–182	Accordingly, four fundamental evaluation criteria were defined to reflect these four key domains of route optimization. The rationale for selecting each criterion and its role in the optimization problem are summarized in Table 1.	Accordingly, four fundamental evaluation criteria were defined to reflect these four key domains of route optimization (Slope/Engineering, Proximity to Roads/Economic, Proximity to Settlements/Social, Proximity to Sensitive Areas/Environmental). The rationale for selecting each criterion and its role in the optimization problem are summarized in Table 1.

Comments 7: [4. Results: Ensure all figures are of sufficient resolution and include clear annotations; several are currently difficult to read.]

Response 7: I appreciate the reviewer’s attention to presentation quality. I acknowledge that several figures, particularly the comparative 2D suitability surfaces and route visualizations (Figures 4–11), may have lost clarity upon submission due to format conversion or scaling.

While the figures were originally generated at a high-resolution standard (600 DPI vector graphics, as per best practice), I will implement the following corrective measures to guarantee maximum readability in the final published form:

1. Increased Readability: I will revise the figures to increase the font size of all annotations, axes labels, and legends to ensure optimal legibility, even when scaled down by the publisher.
2. Vector Submission: I confirm that all final figures will be uploaded in a high-quality, vector-based format (PDF/SVG) to ensure infinite scalability and annotation clarity, fully compliant with IJGI's submission requirements.

 

Comments 8: [5. Discussion: The discussion section should move beyond restating results and instead provide a thoughtful interpretation of their meaning and implications. The discussion should explicitly address whether the study’s objectives were achieved, how the methodology supported those goals, reflect on challenges and limitations, compare findings with previous research (alignment/divergence), and clarify contributions/innovations and future gaps.]

Response 8: I fully agree that the Discussion section must provide a deep interpretation and rigorous analysis that moves beyond descriptive results. The fundamental purpose of this revision was to infuse the Discussion with the necessary theoretical and comparative depth required by the reviewer.

I have significantly RESTRUCTURED and ENHANCED Section 5 to formally establish the theoretical implications of the ISPA framework and address every component of your critique:

1. Interpretation of Methodological Superiority & Objectives (Sec. 5.1): Section 5.1 was RESTRUCTURED to formally establish the theoretical underpinnings of ISPA's success. It now interprets the results by contrasting ISPA’s Spatial Intelligence and Adaptive Behavior against the structural failure of conventional point-wise methods, directly explaining how the methodology achieved its objectives.
2. Comparison to Literature (Divergence): A key comparison statement was ADDED (Lines 809–810) to highlight how ISPA’s deterministic robustness is a significant divergence from the well-documented high sensitivity of other MCDM methods to initial weighting schemes (,), thus extending existing knowledge.
3. Paradigm Shift and Goal Achievement (Sec. 5.2): This section emphasizes that the achieved flexibility (modeling the 'climbing reward') represents a paradigm shift in geospatial optimization, confirming the achievement of unconventional goals.
4. Limitations, Contributions, and Future Gaps (Sec. 5.4): This subsection was EXPANDED to not only list limitations but to frame the future research directions (Dynamic Costs, ML Calibration) as necessary steps to realize the full potential of the GeoAI paradigm introduced by ISPA, thus clarifying the remaining gaps and innovations offered.

[Updated text in the manuscript]:

Line Range	Context	Summary of Revision and Clarification
Lines 790–810 (Sec. 5.1)	Interpretation of Methodological Superiority	RESTRUCTURED/ADDED: This section now provides a deep interpretation of why ISPA succeeds, focusing on the theoretical concepts of Spatial Intelligence and Adaptive Behavior.
Lines 809–810	Sec. 5.1 (Adaptive Behavior subsection)	ADDED NEW COMPARISON TO LITERATURE: This finding demonstrates a key divergence from many traditional MCDM studies, where the optimal result is often highly sensitive to the initial weighting scheme (e.g.,,). ISPA's robustness suggests its GNN-inspired propagation mechanism effectively buffers against the biases and uncertainties inherent in the initial expert judgment.
Lines 811–822 (Sec. 5.2)	Conceptual Flexibility	STRENGTHENED: Discussion on the successful modeling of the unconventional objective ('climbing reward'), confirming goal achievement for the Trekking Trail scenario.
Lines 836–851 (Sec. 5.4)	Limitations and Future Research Directions	EXPANDED: Explicitly acknowledges limitations (static environment, 30m DEM) and details specific future research directions (Integration of Dynamic Costs, ML Parameter Calibration, True 3D Optimization), addressing remaining gaps and future innovation.
All lines in Sec. 5	Pronoun Consistency	REVISED: All plural pronouns were corrected to align with single authorship.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript has been substantially improved, particularly in its theoretical grounding, methodological justification, and analytical depth.

The major conceptual concerns have been effectively resolved. The distinction between the “spatial blindness” of traditional MCDM approaches and the proposed framework is now clearly articulated and appropriately linked to Tobler’s First Law of Geography. The methodological selections—WLC, TOPSIS, and VIKOR—are now well justified within the established literature. The Discussion section has also been meaningfully strengthened; it moves beyond description to provide interpretation, situates the ISPA framework through concepts such as Spatial Intelligence and Adaptive Behavior, and incorporates explicit limitations and future research directions. Technical issues, including pronoun consistency and the revised abstract, have been satisfactorily addressed.

Before final acceptance, two minor points still require attention:

1. Clarification of Originality (Reference 16): Although the methodological explanation has improved, the relationship between this manuscript and the prior conference presentation cited as Reference 16 should be stated explicitly. Please add a brief clarification confirming whether the present manuscript extends, revises, or substantially differs from the earlier work, and affirm the originality of the contribution.

2. Figure Quality Verification: The commitment to submitting high-resolution vector graphics is noted. Final acceptance will depend on confirming that the figures in the final submission—particularly the comparative suitability surfaces—meet readability standards and that all labels and annotations are clearly legible.

Author Response

Responses to Reviewer 3 Comments

1. Clarification of Originality (Reference [16])

Reviewer Comment: Although the methodological explanation has improved, the relationship between this manuscript and the prior conference presentation cited as Reference 16 should be stated explicitly. Please add a brief clarification confirming whether the present manuscript extends, revises, or substantially differs from the earlier work, and affirm the originality of the contribution.

Response: I thank the reviewer for requesting this crucial clarification regarding the originality of the method. I understand the potential for confusion when citing foundational GNN work.

I affirm that the Iterative Score Propagation Algorithm (ISPA) is a novel, original contribution presented for the first time in this manuscript. Reference [16] (Kipf & Welling) is cited solely as the inspiration for the GNN "message-passing principle."

To explicitly affirm the originality and clarify the methodological difference, I have added a clear statement to the Introduction, detailing how ISPA differs structurally from the referenced GNN work:

[Updated text in the manuscript]:

Line Range	Context	Updated Text / Summary of Revision
Lines 77–78 (New Sentence)	...inspired by their fundamental message-passing mechanism [16].	“...inspired by their fundamental message-passing mechanism [16]. Unlike the stochastic, learning-based approach of [16], ISPA is a deterministic and rule-based application designed explicitly for spatial optimization synthesis. “

 

2. Figure Quality Verification

Reviewer Comment: The commitment to submitting high-resolution vector graphics is noted. Final acceptance will depend on confirming that the figures in the final submission—particularly the comparative suitability surfaces—meet readability standards and that all labels and annotations are clearly legible.

Response: I confirm that the previous commitment to high-resolution submission has been fully executed. To guarantee maximum clarity and address the specific readability concerns, the following structural improvements were implemented:

Figure Separation (Structural Improvement): All composite figures showing route alternatives (formerly Figures 5a/b, 6a/b, 7a/b, 9a/b, 10a/b, 11a/b) have been structurally separated into two distinct figures each (e.g., Figure 5 and Figure 6). This change effectively doubles the display area for 2D maps and 3D models, ensuring maximum legibility of labels and routes.

Visual Overhaul: Figures 3, 4, and 8 (Raw Parameters and Suitability Surfaces) have been regenerated and replaced with versions featuring improved color contrast and larger, clearer annotations, resolving any previous visibility issues.

Consistency Check: All internal text citations relating to these figures have been updated according to the new numbering scheme (e.g., old Figure 9 is now Figure 12 and Figure 13).

The final submission package contains high-quality, readable figures that meet the journal's publication standards.

Author Response File: Author Response.pdf