Poster Presentations:
PhD Student Texas A&M University
Comprehensive Community-Level Risk Assessment of Aging Wood Residential Frames Under Seismic Loads
Co-Author: Maria Koliou (Texas A&M University)
Abstract: Communities are facing natural hazards at an increasing rate, and the resulting consequences have become dramatically high. Earthquakes are among these hazards and have recently affected communities, causing severe losses of life and economic damage. Although various studies have examined the impact of earthquakes on wood residential communities and infrastructure to support resilient and robust communities, many overlook age variability in building stocks and, consequently, the resulting differences in performance. In this research, the age variability of buildings—subject to code changes and deterioration that reduce performance and increase consequences—is investigated. Two testbeds in earthquake-prone regions of the San Francisco Bay Area and Portland are considered. These testbeds are assumed to be affected by nearby seismic fault lines associated with the Hayward Fault and the Cascadia Subduction Zone (CSZ), respectively. Two potential earthquake magnitudes, 7.0 and 7.5 moment magnitude (Mw 7.0 and Mw 7.5), are assumed. A combination of the SimCenter tool, Regional Resilience Determination (R2D), and a Python script is used to simulate potential losses. Several decision variables—functional recovery, direct and indirect repair costs, downtime, and fatalities—are used as performance metrics. Results show that variability in loss due to differences in the age of wood residential frames is significant, indicating that targeted mitigation strategies are needed to reduce disparities and enhance community resilience.
PhD Student UC Berkeley
Risk-Informed Earthquake Scenario Selection for Regional Infrastructure Networks Using Disaggregated System Metrics
Co-Author: Matthew DeJong (UC Berkeley)
Abstract: This study develops a framework for prioritizing earthquake scenarios in regional simulations of infrastructure networks. Because medium to high-fidelity regional simulations are computationally expensive, evaluating large scenario catalogs is often impractical. The objective is to define a reduced set of hazard-consistent representative scenarios that preserves the damage and loss patterns that are most relevant to a selected system-level performance metric, for use in higher-fidelity simulations.
The methodology starts from sampled rupture scenarios with occurrence rates and regional spatially correlated intensity measure (IM) maps. A low-fidelity, fragility-based risk calculation is performed across the full set to estimate damage maps and compute a network-level metric (e.g., connectivity loss, recovery time). Exceedance rates of the metric are then disaggregated to identify the main drivers, both in scenario space (rupture location, magnitude, fault system) and in asset space (e.g., assets that belong to a defined corridor). This defines a scenario “failure signature” that describes where consequences happen and how they are distributed across the network. Scenarios are grouped by signature similarity, and representatives are selected to cover the main consequence patterns. Hazard consistency is enforced after selection by assigning weights so the reduced set matches the exceedance behavior of regional IM-map descriptors over the return-period range of interest.
The methodology is demonstrated using BART’s elevated rail network, showing that the selected weighted subset matches the main system-level trends of the full set while keeping clear links between selected events and the structures that drive network performance.
Masters Student University of Waterloo
Holistic Assessment of the Potential Benefits of Residential Seismic Retrofit Program for Canada
Co-Author: Rodrigo Costa (University of Waterloo)
Abstract: Canada is underprepared for the consequences of a damaging earthquake, despite significant risks in British Columbia and the Quebec City-Montreal-Ottawa corridor. National assessments highlight gaps in identifying and prioritizing vulnerable residential buildings, and conventional retrofit evaluations focused on direct loss reduction often fail to justify investment due to high upfront costs. This motivates the need for holistic approaches that integrate physical damages and broader socio-economic consequences to support resilient and equitable mitigation policy.
This research assesses the multi-dimensional benefits of residential seismic retrofits across Canada. Using Monte Carlo sampling, probabilistic damage maps are generated for representative earthquake scenarios from the Canadian National Earthquake Catalogue. Building damage states are simulated under baseline and retrofitted conditions using state-of-the-art seismic risk assessment methodologies, enabling estimation of losses at the community scale.
Beyond structural damage, the framework extends simulated impacts into socio-economic outcomes, including displacement duration, reconstruction delay, long-term household indebtedness, and consumption losses. Comparative analyses with and without mitigation are used to evaluate subsidized retrofit programs, identifying eligibility thresholds and subsidy levels that maximize net benefit. By linking detailed damage simulation with human and economic impacts, this work demonstrates how computational natural hazard modelling can inform data-driven, equity-aware policy design for seismic risk reduction and community recovery.
PhD Student University of Delaware
Linking People to Places: A Synthetic Housing-Household Model for Granular Disaster Risk Assessment
Abstract: Communities assess disaster risk to buildings with growing fidelity, yet most analyses cannot trace consequences to the specific households who live in those structures. We present a framework that links households to buildings by constructing a joint housing-household inventory aligned with the National Structure Inventory (NSI). The approach has three components. First, we generate tract-specific synthetic households from American Community Survey (ACS) microdata and validate fidelity using ``sdmetrics``: Column Shape 0.9985, Column Pair Trend 0.9946, Overall 0.9966, and a Classifier Two-Sample Test (C2ST) of 0.9806, indicating high similarity between synthetic and observed distributions. Second, we train a CLIP-inspired deep contrastive model that embeds household features and NSI housing attributes into a shared latent space to score household-building compatibility. The matcher attains Accuracy 0.927, AUC 0.973, and MCC 0.854 on held-out pairs. Third, we solve a hierarchical multi-objective mixed-integer linear program to allocate households to individual NSI buildings while enforcing building capacities and block/tract-level demographic targets (e.g., total population, elderly share, tenure/occupancy proportions). The resulting inventory enables household-level impact assessment under hazard scenarios by directly propagating modeled building damage to the linked household types, supporting targeted mitigation (e.g., retrofit prioritization) and equitable preparedness planning. This work operationalizes an exposure model that spans physical structures and people, creating actionable inputs for resilience analysis and disaster policy design.
PhD Student Oregon State University
Impact of Spatial Correlation on Population Dislocation Estimates for Cascadia Subduction Zone Earthquakes
Co-Author: Andre Barbosa (Oregon State University)
Abstract: Considering spatial correlation of earthquake ground motions in regional risk assessment is important. Previous studies have shown that neglecting spatial correlation in ground motion residuals can misrepresent uncertainty in regional economic loss estimates, often leading to overestimation of losses from frequent events and underestimation of losses from rare events. Building on prior research, this study presents a methodology for considering spatial correlation for household dislocation following magnitude 8.0 or greater subduction zone events.
A comprehensive case study of Seaside, Oregon, a community facing significant risks from the Cascadia Subduction Zone, is presented. The case study considers: (1) Hazard scenarios from the 2023 USGS National Seismic Hazard Model and disaggregation service, (2) the Goda and Atkinson 2009 spatial correlation model, (3) National Structure Inventory (NSI) building datasets, (4) HAZUS building fragility curves, and (5) the Rosenheim et al. population dislocation model. Results for damage and population dislocation are computed using two community resilience platforms: the Interdependent Networked Community Resilience Modeling Environment (IN-CORE) and the Regional Resilience Determination (R2D) tool. Results indicate that the inclusion of correlation affects the tails of the population dislocation distributions. These tail effects can be relevant for decision-making under uncertainty.