Composite risk indices can produce similar rankings but different priority lists and suggested actions. This study examines these differences across 211 Small Tertiary Planning Unit Groups in Hong Kong using twenty indicators of landslide susceptibility, terrain-based flood susceptibility, exposure and social vulnerability. An equally weighted index is tested through alternative weights, indicator removal and spatial grouping. Evaluation uses 2138 landslide reports from 2015 to 2024. Mixed source dates make the assessment retrospective. Baseline landslide susceptibility has an area under the receiver operating characteristic curve of 0.932 for identifying groups with later reports and 0.641 against an area-weighted background. Across eight alternative designs, rank correlations are 0.871–0.991, but 65.1–95.3% of the reference top 43 groups are retained. Joint weight variation gives median retention of 88.4% (2.5th–97.5th percentiles 67.4–97.7%). Smaller lists and the small physical mitigation category are less stable, and wider weight ranges reduce stability. The reference rule assigns 18 groups to integrated priority, nine to physical mitigation priority, 41 to social resilience priority and 143 to monitoring and land use control. These categories guide local assessment. Flood susceptibility, the composite index and intervention effectiveness require independent outcome validation. Component profiles and stability checks identify decisions needing closer local review.
With the surge of advances in AI and big data, a large number of complex solutions are being developed and proposed to manage and reduce disaster risks and build urban resilience. These solutions include image processing and machine learning for risk assessment and building damage evaluation, as well as early warning systems and enhancements to infrastructure resilience. Although these new technologies can provide more effective and efficient solutions to disaster risk management, they introduce a new challenge in policy crafting and decision-making. Implementation of these solutions requires investments in advance. Each solution reduces part of the disaster risks; some reduce the disaster impact on buildings and infrastructure, some focus on urban socioeconomic indicators, and some solutions build resilience in vulnerable communities. As such, effective decision-making process is vital for selecting among diversity of AI based risk management solutions. The decision-making process should take into account the quantitative evaluation of the proposed AI solutions, as well as the perceptions, characteristics, and priorities of local stakeholders. This paper proposes a novel decision framework by categorizing AI-based solutions for disaster risk management. The proposed framework integrates the knowledge of experts and local decision-makers with quantitative risk assessment results and technology effectiveness evaluations to determine the optimal combination of AI-based solutions for disaster risk management. This framework is applied to a flood scenario to determine the optimal solution for disaster risk management.
Rainfalls with different recurrence intervals, such as the 100-year rainfall, are used by practitioners in stormwater or urban planning and to communicate flood risks to the public. For these statistical terms to be efficient, it is important that they are understood by the intended users, i.e., both practitioners and the public. Through three research questions, we examine the appropriateness of using the 100-year flood or rainfall event for communication of flood risk. The first question treats how flood risk terminology is understood. The second investigates how the perceived risk is affected by communication, and the third looks at what affects a person's understanding of flood risk terminology. Questionnaires were sent to practitioners at Swedish municipalities and house owners in five Swedish municipalities. Both groups showed varied understanding of hydrological return periods, with practitioners demonstrating a better understanding. Additionally, using different expressions to describe flood risk influenced perception, where longer time periods were associated with higher risk perception. Unlike previous studies, understanding of flood risk terminology was quantitatively evaluated, and the focus on house owners and practitioners also included pluvial flood risks. The study concludes that commonly used flood risk terminology is difficult to understand and that communication impacts risk perception.
