Program Results

Introduction to the event

The following is the summary of the research progress and achievements in each research direction during the first term:

Automation and AI-assisted disaster assessment

This research first collected remote sensing images of major domestic and international disasters, ranging from wide-area satellite imagery to localized, multi-angle UAV images. To create datasets for supervised deep learning models, pre- and post-disaster images were manually annotated with the locations of buildings and roads. The damage severity of infrastructure was classified according to international disaster damage assessment guidelines. After integrating and completing the annotations, the dataset was made publicly available, establishing a deep-learning image database for road and building damage identification, and facilitating future research by other organizations.
With the training data collected, appropriate deep-learning model architectures, including object detection, image segmentation, and image classification models, were selected based on the properties of different image sources. These models were used for vertical imagery from satellites and aerial photographs to identify post-disaster building and road damages. Relevant research papers have been published, including eight conference papers (domestic and international) and two journal papers.

Cross-disciplinary disaster interdependency analysis and long-term impact assessment

This research extends previous studies on distributed simulations for disaster events, focusing initially on optimal risk management strategies for critical lifeline systems under seismic events. A review of international post-earthquake service performance quantification methods for various systems and communities was conducted. A network model incorporating the dynamic coupling behavior between social infrastructure systems was developed to simulate earthquake-induced damage and evaluate the resilience of critical lifeline infrastructure.
The analysis considered seismic risk uncertainties under different scenarios, integrating fragility curves, pipeline damage rates, and Monte Carlo analysis. The study explored optimal risk reduction strategies for lifeline systems, aiming to maximize societal resilience. Strategies include seismic retrofitting and post-disaster reconstruction plans. The findings provide seismic risk management strategies tailored to Taiwan's regional lifeline infrastructure, enhancing local and community disaster resilience. Research outcomes include analyses of building, transportation, natural gas, and power network disaster risks. Published works comprise 11 conference papers (domestic and international) and six journal papers.

Disaster Risk Assessment and Taiwan’s Disaster Insurance Studies

This research highlights an exploration of disaster risk from the perspectives of various stakeholders, examining their risk perceptions, preparedness levels, risk tolerance, and willingness to pay. Methods include literature reviews, expert interviews, surveys, and case studies. Integrating social science, psychology, and engineering analysis, the research investigated disaster risks and corresponding strategies across different stakeholders. In addition to surveying public willingness to purchase earthquake insurance, studies were conducted on the willingness of citizens and landlords to adopt building energy retrofitting and seismic strengthening. The research analyzed challenges and critical factors for promoting disaster mitigation policies, aiming to provide references for government strategies to enhance urban disaster resilience. Published works include one conference paper (domestic and international) and two journal papers.