Landslide Risk Assessment of Guangdong Expressway Network Using Random Forest and Complex Network Topological Analysis

Expressways constitute essential lifeline infrastructure supporting regional interconnection and optimizing social spatial patterns. However, geohazards including landslides severely threaten expressway networks, making landslide-triggered risk assessment essential. Conventional static zoning-based evaluations predominantly neglect network functions and fail to effectively quantify the geohazard impacts on network connectivity. To address this gap, this study proposes a landslide risk assessment framework integrating random forest (RF) susceptibility modeling and complex expressway network topological analysis for validation in Guangdong Province, China. Using 13 influencing factors, the RF model produces landslide susceptibility maps with a test set AUC of 0.893, where relief, slope, elevation, geological units, and monthly mean rainfall serve as primary drivers. Furthermore, landslide susceptibility is converted into segment failure probability, and an undirected weighted expressway network graph with 49715 nodes and 45063 edges is established. Betweenness centrality and comprehensive risk index ( R e ) are adopted for quantitative characterization. One random failure and three targeted failure scenarios are developed to simulate variations in global network efficiency and the relative size of the largest connected component. Results indicate that around 22% of core segments govern network connectivity. The R e effectively distinguishes high-risk segments (20% of all edges), among which 6.5% are extremely high-risk links featured by high landslide sensitivity and vital structural importance. Municipal-level risk assessment ranks Huizhou, Yunfu, Heyuan, Meizhou, and Shanwei as the top five cities. These findings provide scientific support for provincial expressway landslide risk mitigation and optimal emergency resource allocation for critical segments.