Rapid assessment of severely affected earthquake areas using mobile signaling data and a random forest approach

This study employed comparative analysis, spatial interpolation, and random forest regression to develop an approach for the rapid assessment of severely affected areas using multiple mobile signaling indicators and multi-temporal analysis. Considering the 2023 Jishishan Ms 6.2 earthquake as an example, this study examined the spatiotemporal dynamics of mobile signaling variations before and after the earthquake. Then the severely affected areas of Jishishan earthquake were identified using the proposed assessment method, with results cross-validated against observed earthquake damage to confirm the method's reliability.

The findings revealed substantial anomalies in mobile signaling data post-earthquake, with varying degrees of responsiveness observed across different seismic intensity zones and indicators. These anomalies were strongly correlated with the actual extent of damage in the affected regions. Additionally, random forest regression was applied to develop seismic intensity evaluation models using mobile signaling data and epicentral distances, further refining the identification of severely impacted zones. In conclusion, the earthquake impact assessment approach using mobile signaling data represents a promising method for the swift and precise identification of severely affected areas. This approach offers a crucial support for post-earthquake emergency response, enhancing both the efficiency and effectiveness of disaster relief efforts.