Climate Vulnerability Assessment of Infrastructure Using Edge‑AI Integrated IoT Systems: A Survey

Approved

1. Relevance and Originality

The paper addresses an increasingly important problem, namely how to assess infrastructure vulnerability under rapidly evolving climate hazards when connectivity, latency, and privacy constraints limit centralized analytics. The focus on edge AI as a resilience enabler is appropriate and timely. The contribution is primarily integrative, bringing together climate risk constructs, IoT architectures, and edge learning techniques into a unified perspective. This synthesis is valuable for readers seeking a cross disciplinary overview, even though the work does not attempt to introduce new analytical models.

2. Methodology

The organization of the survey is logical and thematically consistent, covering sensing, communication, edge learning, digital twins, and governance. The inclusion and exclusion boundaries are clearly stated, which helps manage scope. However, the methodological rigor would be improved by explicitly stating how representative studies were identified and filtered. A brief explanation of literature screening steps, such as databases used or selection criteria beyond publication date and peer review status, would enhance confidence in the completeness of the survey.

3. Validity and Reliability

The discussion is supported by a wide range of authoritative references, including global climate assessments, standards bodies, and recent peer reviewed research. This grounding supports the reliability of the conclusions. The paper appropriately recognizes sources of uncertainty such as domain shift, calibration challenges, and sensor reliability issues. One opportunity for improvement would be a clearer distinction between mature field deployments and experimental prototypes, as both are discussed together and may imply similar readiness levels.

4. Clarity and Structure

The manuscript is clearly written and generally accessible despite the technical nature of the subject. The use of tables to summarize sensor modalities, protocols, and deployment examples is particularly effective. Some sections are dense due to the volume of concepts introduced in close succession, especially in the architecture and evaluation discussions. Introducing short summary statements at the end of these sections could help reinforce key takeaways for readers.

5. Results and Analysis

The analytical value of the paper lies in how it synthesizes reported outcomes from diverse deployments, such as latency reduction, resilience under backhaul loss, and improved decision timeliness. The proposed evaluation scaffold is a strong contribution, as it shifts attention toward decision centric and operational metrics rather than isolated model accuracy. The analysis would be further strengthened by a short comparative discussion of situations where edge AI may offer limited benefit, such as low hazard variability or highly centralized infrastructure systems.