Adaptive Bayesian estimation of earthquake casualties under spatiotemporal dynamics

Abstract

Accurately estimating casualty numbers and injury severity following mass casualty events such as earthquakes is critical for effective emergency response. This study introduces a probabilistic framework for modeling the evolving distribution of earthquake-related injuries across severity categories. Initial casualty probabilities, derived from zone-specific damage assessments, are iteratively updated as new information emerges during ongoing search and rescue operations. In this Bayesian approach, Dirichlet priors are updated with observed injury counts, while the impacts of aftershocks and other secondary hazards are seamlessly incorporated through parameter adjustments. Using both real and synthetic data from the 2023 Türkiye earthquake, we demonstrate how this adaptive framework refines casualty estimates through zone-specific calibration and dynamic integration of aftershock effects.