City-Scale Seismic Risk Assessment Using Optimized Sensor Placement and Sequential Bayesian Updating
Please login to view abstract download link
Earthquakes in densely populated urban regions pose significant challenges for risk assessment and disaster management. Traditional methods often rely on extensive sensor networks to ensure accuracy, which incurs high installation and maintenance costs. To address this limitation, this study introduces an integrated framework for city-scale seismic risk assessment that combines simulation-guided sensor optimization with sequential Bayesian updating. By leveraging physics-based simulations to establish a sparse but efficient monitoring strategy, our approach enables rapid and reliable risk evaluation without the need for densely deployed sensors. The framework begins by establishing an efficient monitoring scheme through a simulation-driven approach. A stochastic Green’s function method is first employed to simulate the propagation of seismic waves from fault zones to the engineering bedrock, thereby generating a comprehensive dataset of surface ground motions. To identify the dominant spatial features from these outputs, proper orthogonal decomposition is applied to the training data, revealing the low-dimensional structure of the seismic response. Guided by these features, we apply a data-driven greedy sparse sensor optimization method\cite{Nakai2021} to strategically select sensor locations. This ensures that the global seismic field is captured with maximum efficiency using a minimal number of sensors. Building upon this optimized monitoring network, the framework implements a dynamic fragility analysis to convert ground motion data into probabilistic estimates of structural damage, following established probabilistic assessment guidelines\cite{Cornell2002}. A key innovation is the implementation of a three-level hierarchical Bayesian model, which groups buildings by structural type while preserving individual characteristics. This model supports sequential updating: as new data are acquired by the sparse sensors, the fragility functions are continuously refined. This mechanism allows for the integration of real-time observations, steadily reducing uncertainty and improving the precision of the risk assessment over time. A case study in an urban area in Japan demonstrates that the proposed framework successfully reconstructs ground motion fields with high accuracy and provides dynamic, detailed insights into structural vulnerability, offering a practical tool for resilient urban planning and decision-making.
