Orienting people-centred disaster shelter planning based on risk assessing with semi-supervised learning

Area of Science:

• Disaster Management
• Geosciences
• Risk Analysis

Background:

• Climate-related disasters, particularly landslides, pose significant threats globally, especially in China's mountainous regions.
• Effective emergency shelter construction and evacuation route design are crucial for public safety and mitigating resident impacts.
• Existing approaches often fail to adequately balance diverse stakeholder needs in disaster preparedness.

Purpose of the Study:

• To develop a novel, people-centric, two-phase methodology for landslide hazard assessment and emergency shelter planning.
• To integrate stakeholder theory, risk economics, and public choice theory into a comprehensive framework.
• To optimize shelter plans that align with government objectives while maximizing resident satisfaction.

Main Methods:

• Phase 1: Landslide hazard assessment model incorporating vulnerable populations using stakeholder theory and risk economics.
• Phase 2: Bi-level multiple objective programming (BLMOP) model to reconcile conflicting government and resident goals.
• Application of a case study in Bazhou Town, China, utilizing field data.

Main Results:

• The hazard assessment model effectively classified landslide points, differentiating risk levels (35.71% high-risk, 64.29% low-risk).
• The BLMOP model generated multiple Pareto-optimal shelter plans, demonstrating a balance between government aims and resident satisfaction.
• The proposed approach proved effective in generating context-specific, optimized shelter solutions.

Conclusions:

• The two-stage, people-focused methodology offers a robust framework for balancing stakeholder needs in landslide disaster management.
• This approach enhances the effectiveness of emergency shelter planning by considering both risk assessment and socio-economic factors.
• The validated case study confirms the practical applicability and effectiveness of the proposed optimization framework.