Poster Presentations:
PhD Student
University of Southern California
Presentation Title: Regional Tsunami Simulation Using R2D Tool
Co-Author: Patrick Lynett
Abstract: This work aims to develop capability within R2D for tsunami impacts including onland structural damage. Existing probabilistic tsunami hazard data for California was gathered, including high-resolution (10-m) maps of tsunami flow depth, speed, and momentum flux at various annual period of occurrence, ranging from ~100 years to ~3000 years. This data was ingested into R2D, such that a user is able to extract a hazard curve at any specified location(s). Fragility functions/curves consist of typical onland structures across all risk category and structure types were organized for the critical assessment related to applicability with infrastructure in California. From within R2D, for a selected location, a hazard curve is extracted from the tsunami database, the appropriate damage or fragility function is selected, and a damage and/or loss analysis is completed.
PhD Student
University of Southern California
Presentation Title: Next-Generation Tsunami Preparedness: A Real-Time, GPU-Accelerated Evacuation Simulator in a Game Environment
Co-Author: Lynett, Patrick
Abstract: "We present an innovative, real-time tsunami evacuation simulator integrating state-of-the-art GPU-accelerated computational fluid dynamics (CFD) using Nonlinear Shallow Water (NLSW) and Boussinesq methods. This is not mere visual effect; our simulation features a highly intricate computational pipeline allowing for accurate water behavior and interactions with buildings and debris. The simulator is uniquely configurable, enabling users to run a range of different tsunami scenarios, each rendered in faster-than-real-time due to GPU parallel processing. The simulator provides a digital twin of real-world cities, generated through high-resolution satellite data, creating an immersive and authentic environment. Within this setting, users engage in character customization options that account for age, physical abilities, health conditions, and custom avatars, making each experience personalized and more relatable. The user interface integrates features like dynamic timers, weather indicators, and map overlays to aid in decision-making. Gameplay realism is enriched through varying terrains and a limited stamina mechanic. Additionally, the educational components are comprised of real-world preparedness tips, emergency contacts, and post-game quizzes. This work significantly advances the field of hazard simulation by coupling state-of-the-art scientific methodologies with tailored educational content. It stands as a powerful tool for individual preparedness training, urban planning, and community resilience against natural hazards, offering broad implications in how societies adapt and prepare for increasing natural disaster risks."
PhD Student
University of Southern California
Presentation Title: VAE as a transfer function to predict onshore hazard curve from offshore information
Co-Author: Patrick Lynett
Abstract: Coastal areas are at risk from various hazards, including tsunamis, storm surges, and sea level rise. These hazards can cause significant damage to property and infrastructure and can also lead to loss of life. In order to mitigate the risks posed by these hazards, it is important to have accurate information about the potential impacts. A critical piece of information is the offshore hazard curve. This curve shows the probability of a tsunami of a given magnitude occurring at a particular location. However, the offshore hazard curve needs to provide information about the onshore impacts of these hazards. Given a particular offshore hazard, the onshore hazard curve shows the probability of inundation at a particular location. The objective of this research project is to develop a transfer function that can be used to convert offshore hazard curves to onshore hazard curves. This transfer function will be developed using a Variational autoencoder (i.e., a deep learning model ). The central idea is to train and validate a model that generates onshore hazard curves based on available offshore hazard data.
As methodology:
A transfer function for offshore to onshore hazard curves represents an exciting hazard assessment and risk mitigation avenue. This project could contribute to more effective disaster preparedness and land-use planning in regions vulnerable to coastal hazards. We look forward to exploring this innovative approach and its potential impact on hazard management beyond California's borders.