Spatial Distribution Characteristics of Marine Economy Based on AI-Assisted Multi-Source Data Fusion and Random Forest Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript presents a study on sustainable production and circular economy evaluation using multi-criteria decision-making (MCDM) techniques. It applies quantitative modeling to evaluate and prioritize sustainability factors across production scenarios, aiming to support more balanced decision-making in industrial sustainability assessment.

The paper is generally well-structured, clearly written, and relevant to the scope of Sustainability. However, the contribution is methodological in form but not conceptually new. Most of the applied methods (e.g., AHP, TOPSIS, or integrated MCDM frameworks) have been widely used in the literature. Therefore, the novelty and validation aspects should be clarified and strengthened before publication.

My feedback on this version of the manuscript:

• The topic is well aligned with sustainability and circular economy evaluation — highly relevant for the journal.

• The paper demonstrates a clear structure linking sustainability indicators, criteria weighting, and decision analysis.
  
  

• The chosen methodological approach (MCDM, possibly combined with weighting schemes like AHP or entropy) is appropriate for the problem type.

• The step-by-step process is presented clearly and is reproducible.
  
  

• The dataset or case example supports the application-oriented focus of the study.

• Indicators and criteria appear comprehensive and relevant.

• The results are logical and generally well interpreted.

• Graphical presentation (figures/tables) supports understanding.

There are some Weaknesses and lack of explanations:

• The study heavily relies on standard MCDM tools, and the innovation is limited to the application context rather than methodological contribution.

• The scientific gap is vaguely defined — the introduction needs to specify what is missing in previous MCDM-based sustainability assessments and what this work uniquely adds.

  • Lacks justification for why the specific MCDM technique was chosen over others (e.g., why TOPSIS or AHP instead of PROMETHEE or VIKOR).

  • It is unclear how weights were derived — subjective expert judgment, data-driven normalization, or hybrid?

  • There is no mention of sensitivity analysis to test the stability of rankings under changing weights, which is crucial in MCDM validation.

  • The data sources are not clearly described (primary, secondary, or simulated).

  • The scale of data (sample size, geographic coverage) is missing. Without this, it is difficult to evaluate the model’s representativeness.

the literature review provides adequate coverage of sustainability and MCDM topics but lacks recent (2023–2024) studies that use hybrid or AI-based decision-support frameworks.
To modernize the review, consider citing works on:

• Machine Learning/Artificial Intelligence for Sensor Data Fusion–Opportunities and Challenges — 10.1109/maes.2020.3049030

• Toward Digital Twin of the Ocean: From Digitalization to Cloning — 10.1007/s44295-023-00003-2

• A Review of Artificial Intelligence in Marine Science — 10.3389/feart.2023.1090185

• Computational Engineering Based Approach on Artificial Intelligence and Machine Learning-Driven Robust Data Centre for Safe Management — 10.53759/7669/jmc202303038

• Using AIS Data to Analyze and Optimize Vessel Operations for Offshore Wind Farm Maintenance — 10.1007/978-3-032-02102-1_31

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

First of all, I would like to thank the journal editor for giving me the opportunity to review the manuscript.

To address the lack of clarity regarding the spatial distribution characteristics and driving mechanisms of the marine economy, this paper proposes an analytical method integrating artificial intelligence (AI) technology and multi-source data. Through kernel density estimation, spatial autocorrelation analysis, buffer analysis, and combining Spearman correlation and random forest regression, it systematically analyzes the spatial pattern and influencing factors of China's marine economy from 2013 to 2023. The results show that China's marine economy exhibits a spatial gradient characteristic of "hot in the south and cold in the north, strong in the east and weak in the west," with coastal zone economy being the dominant driving force. Key influencing factors include the proportion of non-agricultural industries, offshore distance, and per capita marine GDP. Overall, this study provides empirical evidence for the allocation of marine spatial resources and the optimization of industrial structure, and promotes the application of AI in marine economic geography research. However, some shortcomings remain in terms of content and format, mainly including the following aspects.

Suggestions for modification:

1. The timeliness and representativeness of data need to be strengthened. It is suggested to introduce higher frequency or real-time data (such as night lighting, AIS ship trajectory, etc.) to enhance the timeliness and representativeness of the research.
2. The analysis of driving mechanism can be further deepened. Although the study identified the key influencing factors, the discussion on their action paths and interaction mechanisms was relatively simple. For example, the causal relationship between the proportion of non-agricultural industries and the marine economy has not been explained in depth.
3. Policy recommendations lack specificity and regional differences. It is suggested that more operational policy guidance should be put forward by regions (such as Bohai Sea, Yangtze River Delta, Guangdong, Hong Kong and Macao, etc.) to enhance the practical guiding value of research results.
4. The basic elements of the map in Figure 1 are missing, so it is suggested to supplement the basic map elements such as the north arrow, legend, latitude and longitude grid, and add labels to each province.
5. The innovation of method integration is insufficient. Despite the emphasis on "AI technology and multi-source data fusion", the specific description of how AI is applied to spatial analysis (such as model selection, training process, verification results, etc.) is not sufficient.

Accept after minor revision

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Review Report

This manuscript offers a comprehensive spatial analysis of China’s marine economy using AI, multi-source data fusion, and GIS-based methods. The topic is timely and aligns well with Sustainability’s focus on sustainable coastal management. The study is data-rich and uses an impressive mix of spatial and statistical tools, providing valuable insights into regional disparities and marine-economic patterns. However, the analysis is largely descriptive, and the AI component is basic, limited mainly to data handling and Random-Forest regression. Key methodological details including data validation, model parameters and uncertainty treatment are missing which limits scientific rigor and reproducibility.

Overall, the paper has clear potential but requires major revision to clarify methods, strengthen the analytical depth and better demonstrate how AI meaningfully enhances understanding of marine-economic spatial dynamics. Below are my detailed observations and suggestions:

Abstract

The abstract clearly outlines the study’s scope, data and methods using AI, multi-source data fusion and GIS tools to assess China’s marine economy. It summarizes results well but feels too descriptive. The novelty of the AI integration is overstated, as only basic machine-learning tools are applied. Adding one or two key quantitative findings and a line about policy relevance would make it stronger.

Introduction

The introduction offers good background and context but is overly long and historical. It spends more time on policy and history than on scientific gaps. The literature review lacks up-to-date global research on AI-GIS methods. The authors should shorten background material, include more recent studies, and clearly state the research question and contribution at the end.

Research Methods and Data Sources & Evaluation Index System and Research Methods

The methodology is clear and well-structured, combining spatial statistics and machine learning. However, the AI use is limited to data scraping and Random-Forest regression. Key details on sample size, preprocessing and validation are missing, and provincial-level data may not suit fine-scale tools like Kernel Density. The authors should clarify data structure, model parameters, and validation steps. Figure 1 should include clear geographical boundaries, labeled provinces, and map elements (scale, north arrow, projection), while Figure 2 requires higher resolution to ensure readability and accurate interpretation.

In addition, Section 2 (Research Methods and Data Sources) and Section 3 (Evaluation Index System and Research Methods) overlap considerably, both describing datasets and analytical procedures. The authors should merge or clearly differentiate these sections to avoid duplication and improve structural clarity. This will streamline the paper and avoid redundancy.

Results and analysis

The figures and structure are clear, and the (south-hot, north-cold) spatial trend is convincing. Still, the section is too descriptive and lacks quantitative detail. Figures need clear legends and units, and uncertainty or error values should be added. Random-Forest results show variable importance but not model accuracy. A few tables with key numbers would make this section more credible.

Discussion and Policy Recommendations

This section connects results to policy well and fits Sustainability’s focus. The recommendations are clear and practical. However, the discussion repeats results rather than comparing them to other studies. The authors should better link their AI findings to actual policy tools and briefly reference global marine-planning frameworks for stronger context.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript explores the spatial distribution characteristics of China’s marine economy (2013–2023) through the integration of artificial intelligence (AI) and multi-source data fusion. The topic is highly relevant to sustainable coastal development and spatial governance, aligning well with China’s “marine power” strategy. The work provides extensive spatial analysis using established tools (KDE, Moran’s I, Getis–Ord Gi*, buffer analysis) and supplements them with a Random Forest model to assess influencing factors.

However, the study is mainly descriptive, with limited methodological transparency and modest theoretical innovation. The “AI + multi-source fusion” concept is under-specified, and several analytical steps lack reproducibility. The results confirm well-known spatial gradients without offering new causal or predictive insights. Overall, the manuscript is suitable for an applied sustainability journal after major revisions aimed at improving methodological rigor, data transparency, and scientific framing.

Specific Comments

Abstract

• p.1, lines 10–26: The abstract claims “AI technology with multi-source data fusion” but does not specify what AI methods were applied (e.g., model types, training process). Clarify whether AI was used for data collection, prediction, or feature weighting.
• p.1, lines 19–23: The phrase “driving mechanisms of China’s marine economy” implies causal inference; rephrase as “factors influencing spatial differentiation.”

Introduction

• p.1-2, lines 33–51: The introduction reads primarily as a policy and historical overview. Condense national strategic background (lines 33–41) and emphasize the methodological innovation (lines 44–50) more clearly.
• p.2, lines 47–50: Explicitly state the study’s research questions or hypotheses; the stated purpose (“analyzing spatial distribution patterns and driving mechanisms”) remains too broad and descriptive.
• p.2-3, lines 77–107: The historical overview of China’s maritime development (pre-Qin to modern era) is informative but dense. Add a timeline figure showing main dynastic periods (pre-Qin, Han, Ming, Qing, Modern) and key maritime milestones to clarify the evolution of China’s marine economy.

Research Methods and Data Sources

• p.4, lines 137–155: The data-fusion process is described in generic terms (“AI web-scraping,” “Excel data matrices”). Provide concrete details on variables, sources, and quality-control steps.
• p.6-7, Table 1: Several indicators (e.g., “per capita urban and rural construction land area”) lack units and definitions. Add data type, source, and normalization method.
• p.8, lines 237–260: The “AI data preprocessing” section would benefit from a diagram showing the multi-source data pipeline.
• p.9, Eq. (2–3): For Moran’s I, specify the spatial weight matrix type (contiguity vs. distance-based) and distance threshold used.
• p.10, lines 333–337: The Random Forest workflow (R package) is introduced but without reporting parameters (number of trees, max depth, feature sampling). Add these to ensure reproducibility.

Results and Analysis

• p.12, Figure 4: The kernel density maps are descriptive. Include quantitative statistics (mean density, standard deviation, area share) to support qualitative interpretations.
• p.13, Table 2: Global Moran’s I values are significant, but no spatial correlogram or Local Moran (LISA) cluster map is provided. Adding local autocorrelation results would strengthen spatial interpretation.
• p.15, Figure 6: The Getis–Ord Gi* hotspot results should be accompanied by explicit Z-score thresholds and classification criteria used to define hot-, sub-hot-, cold-, and sub-cold-spots.
• p.16–17, Figure 7: The buffer-zone analysis claims that “wider buffers exhibit stronger economic activity.” Provide numerical metrics (e.g., GDP or intensity per buffer zone) instead of purely descriptive statements.
• p.18, Figure 8: “Overall spatial capability” is discussed, but the composite-index computation method is not given. Define the weighting or normalization procedure used to aggregate indicators.

Statistical and Machine-Learning Analysis

• p.19, Figure 9: The Spearman correlation coefficients are generally below 0.3, indicating weak relationships. Discuss the implications—these low values suggest limited explanatory or predictive power for individual variables.
• p.20, Table 3: The Random Forest section reports R² = 0.926 but does not describe the training/test split ratio or validation method (e.g., k-fold CV). Add cross-validation details and uncertainty measures such as confidence intervals or error bars.
• p.21, Figure 11: The feature-importance weights are highly imbalanced. Clarify whether they were derived from normalized importance, permutation importance, or Gini-based metrics, and explain how scale effects were mitigated.

Discussion and Policy Implications

• p.22–24: The discussion emphasizes policy recommendations but lacks quantitative or literature-based comparison to prior spatial-economic models. Relate findings to existing marine economic-geography studies (e.g., comparative regional spatial models or coastal zone development frameworks).
• p.22, lines 695–699: The call for industrial cooperation zones is policy-oriented; link it explicitly to spatial evidence from your results (e.g., Moran’s I clustering, hotspot regions) to justify this recommendation empirically.
• p.23, lines 730–733: The proposal for “enclave economy” cooperation zones (North Gulf region) is forward-looking but again lacks grounding in the quantitative findings. Clarify how the identified low-density or coldspot areas justify this recommendation.
• p.24, lines 739–747: The suggestion of “blockchain-based fishing quota trading” is interesting but beyond the study’s empirical scope. Consider shortening or framing as potential future research rather than a concrete policy proposal.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

• The manuscript is well-structured and addresses a relevant topic for Sustainability, focusing on spatial patterns of China’s marine economy with a combination of spatial statistics and machine-learning-based importance analysis.

• The overall methodology is appropriate. Kernel density, Moran’s I, Getis–Ord Gi*, and buffer analysis are correctly applied, and the Random Forest model is a reasonable choice for exploratory factor importance.

• The contribution framed as “AI + multi-source data fusion” is somewhat overstated. The AI component is mostly limited to data collection and a standard RF model. I suggest slightly moderating the wording for accuracy, but this does not affect the scientific soundness.

• Some methodological details are insufficiently explained, especially regarding the data fusion workflow, indicator construction, and RF model specification. Please add a brief description of the tools/packages used, the construction of the composite indices, and key RF hyperparameters.

• The discussion and policy recommendations are informative but somewhat lengthy. Streamlining this part would improve focus and readability.

• Several required MDPI sections are missing or incomplete (Data Availability, Funding, Conflicts of Interest, Author Contributions). Please add these according to MDPI standards.

• English language is generally understandable, but the manuscript would benefit from minor polishing to fix spacing artifacts, hyphenation issues, and long sentences.

• There are citation numbering inconsistencies that must be corrected.
  The following in-text citations do not match the first authors of the corresponding references:

  • Tahat [41] → Reference 41 is Blasch

  • Sun [42] → Reference 42 is Chen

  • Zhang [43] → Reference 43 is Song

  • Alahmari [44] → Reference 44 is Senthilkumar

  • Kowalski [45] → Reference 45 is Baghizadeh
    Please correct either the names in the text or the numbering.

After addressing the above issues, the manuscript will be suitable for publication.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript has been clearly improved. The structure is more coherent, the methods are better explained and the results now include stronger quantitative evidence and clearer figures. The revisions address the previous concerns and the paper is now well-prepared.

Author Response

  Comments 1：The manuscript has been clearly improved. The structure is more coherent, the methods are better explained and the results now include stronger quantitative evidence and clearer figures. The revisions address the previous concerns and the paper is now well-prepared.

  Response 1：We appreciate the reviewer’s recognition that the previous concerns have been fully addressed and that the paper is now well-prepared. Your constructive comments greatly contributed to improving the overall quality of the manuscript.