Dynamic assessment of earthquake entrapped occupant risk using building-scale population modeling: A case study of Yuxi City, China

Rapid assessment of entrapped occupants is vital for improving post-earthquake rescue efficiency and reducing casualties. Conventional approaches based on administrative units or grid cells cannot capture the fine-grained, dynamic distribution of population at the building level. To address this gap, this study proposes a building-scale framework for entrapped occupant risk assessment in Yuxi City, Yunnan Province, China. The framework integrates four components: building function identification, population estimation, indoor occupancy correction, and entrapped occupant assessment. A hybrid method combining Areas of Interest (AOI)- and Points of Interest (POI)-based direct identification, spatial similarity inference, and Kernel Density Estimation (KDE)-based recognition was developed to improve building function classification, especially in areas with sparse POI data. A random forest model was then used to derive population influence weights for different building function types, achieving building-scale population allocation with an overall error below 5%. By incorporating time-dependent indoor occupancy rates and a building vulnerability matrix, a building-level assessment model for entrapped occupants was established. Seventy scenarios were designed based on five seismic intensity levels (VI–X) according to the Chinese Seismic Intensity Scale, seven time periods, and two day types to simulate the spatiotemporal distribution of entrapped occupants in Hongta District. The results show that building function influences entrapment risk: healthcare and educational buildings exhibit risks 3–5 times higher than residential buildings. Public buildings present higher risk during 09:00–18:00 on working days, while commercial and recreational buildings show increased risk on non-working days. Residential buildings remain at high risk at night. The proposed method enables dynamic entrapped occupant risk estimation at individual-building resolution and supports post-earthquake emergency rescue in earthquake-prone regions.