Assessment and zoning of seismic landslide hazards in Sichuan, China, using a SCM-ANFIS model under different ground motion

• Published in: Bulletin of engineering geology and the environment Vol. 84; no. 4; p. 184
• Main Authors: Wang, Jie, Xu, Chong, Xie, Zhuojuan, Li, Yu, Zhang, Lifang, Lv, Yuejun
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2025; Springer Nature B.V
• Subjects: Adaptability; Algorithms; Clustering; Earth and Environmental Science; Earth Sciences; Earthquake magnitude; Earthquakes; Emergency preparedness; Foundations; Geoecology/Natural Processes; Geoengineering; Geological hazards; Geology; Geotechnical Engineering & Applied Earth Sciences; Ground motion; Hazard assessment; Hydraulics; Landslides; Landslides & mudslides; Machine learning; Movement; Nature Conservation; Neural networks; Original Paper; Parameters; Regression analysis; Sampling methods; Seismic activity; Seismic hazard; Shaking; Support vector machines; Topography; Zonation; Zoning; China; Sichuan China; SCM-ANFIS model; Landslide hazard; Seismic zoning; Sichuan Province; Seismic landslide
• ISSN: 1435-9529; 1435-9537
• DOI: 10.1007/s10064-025-04172-8

The distribution and intensity of seismic landslides are directly influenced by factors such as earthquake magnitude, epicentral distance, and focal depth; therefore, evaluating seismic landslide hazards should be based on a comprehensive assessment of seismic hazards. In this study, we focused on Sichuan Province, an area characterized by a high susceptibility to significant seismic activity and a propensity for landslides triggered by earthquakes. Initially, we used the potential source model, activity parameters, ground motion prediction equations, and calculation methodologies outlined in the current GB18306-2015, “Seismic Ground Motion Parameter Zonation Map of China” to determine the distribution of peak ground acceleration (PGA) for basic ground motion, rare ground motion and extremely rare ground motion (with a 50-year exceedance probability of 10%, 2% and 0.5%) in Sichuan Province. We then applied a hybrid machine learning model that combines high predictability and tractability, known as the Subtractive Clustering Method-based Adaptive Neural Network Fuzzy Inference System (SCM-ANFIS). The model used earthquake landslide databases from Wenchuan, Lushan, Jiuzhaigou, and Luding, along with 12 relevant factors, including topography and seismic geology. We established a coseismic landslide hazard assessment framework and evaluated seismic landslide probabilities under three levels of ground shaking in Sichuan Province, resulting in the creation of a hazard zoning map. Finally, we assessed the sampling methodology, adaptability, and limitations of the model while also exploring its potential applications. This research can significantly improve disaster prevention and management and inform infrastructure development in Sichuan Province. Future efforts will focus on enhancing data breadth and precision.