Natural Hazards Review

Computer simulation is increasingly being used by emergency planners as a tool to improve disaster response given that it can model real-world scenarios, such as earthquakes. Although there has been an increase in simulation research focused on disaster response, much of this literature is from disparate fields and across disaster scenarios. To bridge this gap, this paper surveys simulation-based models for postearthquake response from the year 2000 onward. Advantages of simulation over closed-form statistical models are discussed. Three main subproblems in postearthquake response models are explored: (1) service distribution (e.g., food, water), (2) infrastructure restoration (at the building and transportation system level), and (3) emergency healthcare provision (i.e., casualty treatment). Additionally, the underlying simulation-optimization model in its canonical form and numerical results from case studies are presented to illustrate the nature and size of the problems encountered in the real world. Thereafter, gaps in the literature are identified keeping in mind the objectives of FEMA’s recent strategic plan, which focuses largely on equity and resilience. This paper revealed key misalignments between policy priorities and existing literature—specifically, that there is a gap in equity-based simulation modeling.

Natural disasters have caused significant economic losses and casualties. Obtaining detailed disaster information and understanding public opinion during disasters are crucial for devising effective policies and ensuring timely disaster responses. With the widespread use of social media, it has become an important channel for extracting disaster information. However, accurately obtaining and revealing public opinion from social media remains challenging. This study combines the biterm topic model and support vector machine to analyze topic features. Additionally, sentiment features are analyzed using the Generative Pre-trained Transformer-3.5 model. These techniques are employed to build a social media-based flood information mining model capable of detecting the spatiotemporal distribution of public sentiment and discussion topics, including significant events impacting public sentiment. Using the 2022 Yingde flood as a case study, we explored the evolutionary patterns of public opinion on floods across three dimensions: time, space, and content. The study also explored the correlation between flooding and public opinion through geographic visualization and statistical analysis. The results indicated a precision of 89.2% and 80.2% for topic and sentiment classification, respectively. Temporally, the public response to flooding was primarily concentrated during heavy rainfall and flooding, varying with disaster severity. Furthermore, significant events or statements by public figures can greatly influence public responses to flooding. Spatially, the public response to flooding focused mainly in major urban areas and severely affected regions. In terms of content, a strong correlation was revealed between sentiments, topic distribution, and the disaster scenario. The findings can be used to analyze disaster conditions and public opinion in depth, and as a supplement of existing methods of extracting disaster information, it can enhance situational awareness for disaster emergency management and provide a reference basis for emergency relief efforts.