Modelling spatial correlation in earthquake-induced damage and its impact on regional loss estimation

• Main Authors: Nguyen, Michele, Loos, Sabine, Lallemant, David
• Other Authors: Asian School of the Environment
• Format: Conference or Workshop Item
• Language: English
• Published: 2023
• Subjects: Engineering::Civil engineering::Spatial information/surveying; Science::Mathematics::Statistics; Spatial Correlation; Random Fields; Aggregate Seismic Loss
• Online Access: https://hdl.handle.net/10356/169433; http://www.icossar2021.org/
• Institution: Nanyang Technological University

While correlation in ground motion intensity and its effect on earthquake damage has been well-documented, there is a widely held yet insufficiently examined view, that there are other sources of spatial correlation in damage not accounted for. To provide further evidence and illustrate impacts on regional losses, we develop a spatial ordinal damage model. This involves adding Gaussian random fields at a latent level and using data from multiple building types to separate the sources of spatial correlation in the absence of sensor data. Unlike previous approaches which require two fitting steps, this novel model is fit once using maximum likelihood estimation. As illustrated using data from the 2010 Haiti earthquake, the modelled spatial correlation leads to greater regionalization of the latent variable mean surface and shifts in damage state distributions. This highlights the increased probabilities of extreme regional loss due to the additional factors for spatial correlation in damage.