Conflict-Based Models for Real-Time Crash Risk Assessment: A State-of-the-Art Review

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Conflict-Based Models for Real-Time Crash Risk Assessment: A State-of-the-Art Review

The manuscript addresses a highly relevant and timely topic in transportation safety engineering, namely the use of conflict-based models for real-time crash risk assessment. The subject matter is important, and the paper successfully highlights why traditional crash-based models are limited for real-time applications, particularly because crash data are rare, delayed, and inherently stochastic. The attempt to synthesize the literature through five modelling paradigms—statistical and regression-based models, Bayesian frameworks, extreme value theory models, machine learning approaches, and hybrid models—is valuable and gives the paper a clear overall structure. In that respect, the manuscript has the potential to make a useful contribution to the field. However, in its present form, the study is not yet sufficiently rigorous to be considered a strong state-of-the-art or systematic review article, and I therefore recommend major revision.

My main concern relates to the review methodology. Although the manuscript states that PRISMA guidelines were adopted and that the search was conducted in Scopus, Web of Science, and ProQuest, the methodological description remains too limited to ensure transparency and reproducibility. The paper reports the search string, the publication window, and the number of studies retained, but it does not provide enough detail regarding the full eligibility criteria, the exact search procedure in each database, the screening logic, the reviewer workflow, or the way disagreements were resolved during study selection. A review that presents itself as systematic should do more than report the number of records found and retained; it should demonstrate a transparent and reproducible protocol. In its current state, the methodology section is too brief and should be substantially expanded.

A related issue is that the search strategy itself appears rather narrow for a paper that claims to provide a comprehensive synthesis of conflict-based real-time crash risk modelling. The search string based on “traffic conflict OR conflict-based” and “real-time OR short term” may not be broad enough to capture the full spectrum of relevant studies in this domain. Many papers in this literature may use adjacent terms such as surrogate safety, near-miss analysis, dynamic safety evaluation, short-term crash prediction, trajectory-based risk prediction, or online risk assessment without necessarily using the exact terms adopted by the authors. For that reason, the manuscript should either justify more convincingly why the selected query was sufficient or revise the search strategy to demonstrate that the literature base was captured more comprehensively.

In terms of content, the manuscript is readable and well organized, but the synthesis remains largely narrative and descriptive. The paper does provide a useful overview of representative studies under each modelling paradigm, yet it does not sufficiently move from description to analytical synthesis. A stronger review would not only describe model families, but also systematically compare them across concrete dimensions such as traffic environment, data source, conflict indicator, prediction horizon, validation design, computational burden, and operational readiness. At present, the discussion gives the reader a general sense of the field, but it does not yet provide the level of structured evidence mapping that would be expected from a high-quality state-of-the-art review.

The manuscript also needs to sharpen its claim of novelty relative to prior review papers. The Introduction correctly notes that the literature is fragmented and that previous work has reviewed traffic conflict modelling and real-time crash prediction. However, the present manuscript does not yet explain with enough precision what is genuinely new about this review. The authors should state more explicitly whether the originality of the paper lies in its exclusive focus on real-time conflict-based modelling, in its model-centric structure, in its treatment of uncertainty and temporal dynamics, in its focus on hybrid models, or in its distinction between conflict frequency and conflict severity. At the moment, the novelty is implied but not sufficiently demonstrated.

One potentially original and interesting feature of the paper is the distinction drawn between conflict frequency modelling and conflict severity modelling. This is a promising conceptual lens, especially in the cross-comparison section and in the related comparative table. However, the manuscript does not yet develop this distinction with enough theoretical clarity. It would be useful for the authors to define more precisely what they mean by frequency and severity, to explain whether these are complementary or overlapping concepts, and to show more systematically how each modelling family contributes to one or both of these dimensions. As currently written, the argument is interesting, but it remains somewhat underdeveloped and would benefit from deeper conceptual grounding.

The classification into five modelling paradigms is practical and helps organize the literature, but the boundaries between categories are not always fully clear. Several studies cited in the Bayesian and EVT sections also have hybrid characteristics, and some of the hybrid studies appear to overlap conceptually with earlier categories. For this reason, the manuscript should clarify how category assignment was performed and whether the classification is meant to be mutually exclusive or simply heuristic. Without such clarification, the structure remains useful for reading purposes but somewhat subjective from a methodological standpoint.

The paper also makes some evaluative claims that should be stated more cautiously. For example, the manuscript suggests that hybrid models represent the most advanced and most promising approach for real-time crash risk assessment. While this may be a reasonable interpretation, the review does not present sufficiently systematic comparative evidence to support such a strong conclusion. Methodological sophistication does not automatically imply operational superiority. Real-time applicability also depends on data availability, interpretability, computational efficiency, calibration burden, and integration with traffic-management systems. The authors should therefore moderate such claims and distinguish more clearly between theoretical promise, predictive performance, and practical deployability.

The discussion of validation, transferability, and deployment is relevant, but it should be more deeply integrated into the central analysis of the paper. The manuscript correctly notes that many studies are site-specific and that external validation remains limited. It also identifies scalability and real-time deployment as open challenges. These points are important, particularly for civil and transportation engineering practice, but they currently appear more as concluding observations than as core analytical findings. The review would be significantly stronger if it explicitly compared how different paradigms perform with respect to cross-site generalizability, robustness to data shifts, dependence on specific sensing technologies, and real-world computational feasibility.

There are also several structural and editorial issues that should be corrected. One clear problem is that subsection 5.1.2 repeats the heading of subsection 5.1.1, even though the actual content of that subsection concerns rare and extreme events rather than standardization in conflict definitions. In addition, the manuscript contains placeholder journal metadata on the first page, including generic editorial information that should obviously not remain in the submission. There also appear to be inconsistencies in the internal references to tables, since the text refers to tables that are not clearly visible in the provided manuscript excerpt. Minor proofreading is also needed, as there are some typographical, punctuation, and formatting irregularities throughout the paper.

Overall, I believe the manuscript addresses an important topic and has a solid conceptual foundation, but substantial revision is needed before it can be considered for publication. The review protocol must be reported with greater methodological rigor, the evidence synthesis must become more analytical and systematic, the paper’s novelty relative to prior reviews must be clarified, and the conceptual discussion—especially regarding conflict frequency and severity—should be strengthened. With these improvements, the manuscript could become a useful contribution to the literature on proactive and data-driven road safety analysis.

Author Response

Please see attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This review addresses a timely topic and offers a useful model-centric categorisation of conflict-based approaches for real-time crash risk assessment. 

• However, it remains largely descriptive rather than analytical, with limited critical comparison of model performance, data requirements, and real-world development evidence. The paper would benefit from a clearer review methodology, quantitative synthesis (e.g., benchmarks), and stronger discussion on validation, transferability, and practical implementation challenges.
• So I guess adding research questions or review objectives would strengthen the paper’s structure and impact.
• It is recommended to add a graph or a sketch in section 2, to show the methodology process. 
• In tables 1 and 2, I see a nice comparison in general and highlight some themes, while I think you can elaborate more and deepen the analysis by focusing on sample size, software, implementation, and the limitation model in a more specific way, even by numbers. 
• I suggest adding more descriptive analysis on chosen papers, the field of study, and more themes, like impacts. 
• You may show the outcomes in figures as well. 
• Conclude with actionable guidance by formulating concrete recommendations for researchers and policymakers and illustrating the methodological workflow with a schematic figure. 
• Clarify the PRISMA review protocol with explicit inclusion/exclusion criteria and add a flow diagram to improve transparency and reproducibility of the review process.
• Strengthen the analytical depth by adding quantitative comparison (sample size, data type, software/tools, validation metrics, and limitations) across studies rather than thematic description only.

 

 

Author Response

Please see attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Editor,

Thank you for the opportunity to review the revised manuscript.

Based on the revisions made in response to the reviewers’ comments, I am satisfied that the necessary corrections have been adequately addressed. Therefore, I recommend that the manuscript be accepted for publication.

I would also like to thank the authors for their careful revisions and constructive responses to the comments.

Sincerely,

Reviewer 2 Report

Comments and Suggestions for Authors

After reviewing the paper for a second time, I can see that the authors have addressed almost all of my comments.

My only remaining suggestion concerns Figure 1 (Methodological Framework). I believe the position of the title should be adjusted for better clarity and presentation.