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

Approved

1. Relevance and Originality

The paper addresses a problem of clear practical importance by examining how edge AI and IoT can support climate vulnerability assessment for infrastructure under time sensitive and connectivity constrained conditions. The emphasis on standards led interoperability and governance aligned deployment is a notable strength. Originality stems from the way climate risk concepts are operationalized within edge centric data flows and decision pipelines, rather than from algorithmic novelty. This positioning is appropriate for a survey intended to guide applied research and deployment.

2. Methodology

The survey is carefully scoped and organized around a coherent system lifecycle, from sensing to decision support. The articulation of inclusion criteria helps keep the discussion focused. However, the paper would benefit from a more explicit description of the literature review process, including how sources were identified, screened, and categorized. Clarifying whether the review is exhaustive or illustrative would help readers interpret the breadth of coverage. Additionally, some sections would gain clarity by explicitly stating assumptions about deployment scale, asset heterogeneity, and environmental conditions.

3. Validity and Reliability

The paper demonstrates strong alignment with authoritative sources and standards, including IPCC assessments and NIST guidance, which supports the validity of its claims. The discussion of calibration, uncertainty, and governance is particularly well grounded. Reliability could be further improved by more clearly separating evidence from inference when summarizing reported performance outcomes. In several instances, results from different contexts are presented together, which may obscure differences in operating conditions, sensor density, or evaluation protocols.

4. Clarity and Structure

Overall organization is strong, with a logical progression from conceptual framing to technical architecture and application domains. Tables and figures are well chosen and help condense complex information. At times, the narrative becomes dense due to the simultaneous introduction of multiple standards, protocols, and learning approaches. More frequent signposting or short recap statements would improve navigability, especially for readers who are less familiar with IoT or edge computing ecosystems.

5. Results and Analysis

The paper effectively synthesizes reported benefits of edge AI, such as reduced latency, improved continuity during backhaul loss, and better alignment with privacy and governance requirements. The evaluation scaffold proposed is a valuable contribution and encourages more decision relevant benchmarking. To strengthen the analysis, the authors could include a brief discussion on operational tradeoffs, such as maintenance overhead, lifecycle costs, and the human factors involved in interpreting edge generated alerts within asset management workflows.