Hazard assessment for typhoon-induced shallow landslides based on rainfall thresholds and physical modeling

Typhoon-induced shallow landslides have caused significant economic losses and casualties in China's coastal regions. Accurate prediction and hazard assessment of typhoon-induced landslides are crucial for effective geohazard prevention and management. However, providing accurate hazard evaluation remains challenging due to limited data on rainfall triggers and relevant geological parameters. Therefore, our study integrates the effective rainfall model and the probabilistic physical model TRIGRS to analyze the early warning of regional shallow landslides. In this study, we selected Daoshi Town, in Zhejiang Province of China, which was heavily impacted by Super Typhoon Lekima on August 10, 2019. To find out the distribution and regularity of landslides after typhoon rainfall, we identified a total of 190 shallow landslides through field surveys and remote sensing interpretation. The soil thickness of the study area was simulated using the random forest algorithm based on the soil thickness dataset from the field survey. Rainfall characteristics and thresholds were established using an effective rainfall model that accounts for the 6-hour rainfall on the day of analysis and the cumulative rainfall over the preceding three days. To assess slope stability under different rainfall scenario, TRIGRS was employed, considering key parameters of different soil types such as cohesion and internal friction angle. The results indicate that 90% of the landslides occurred in areas classified as highly unstable. Validation using landslide data from 2020 and 2021 showed that 81% of new landslides occurred in unstable areas, demonstrating the reliability of the proposed early warning approach. It shows that our results are reliable and can provide reference for the hazard assessment and management of typhoon-induced shallow landslides in coastal regio.