Early warning models for rainfall-triggered clustered landslides: Advances, challenges and future perspectives

Against the backdrop of global climate change and the increasing frequency of extreme rainfall events, rainfall-triggered clustered landslides have become a major concern in mountainous disaster risk management due to their sudden occurrence, regional clustering, and high destructive potential. This paper systematically reviews the current development and technological advances in early warning models for rainfall-triggered clustered landslides. Particular emphasis is placed on the theoretical foundations and application characteristics of rainfall threshold models, hydrological and physically based models, data-driven models, and physics–data-driven coupled models. The critical role of multi-source monitoring technologies in regional landslide early warning systems is also examined. The review indicates that empirical statistical threshold models, characterized by high computational efficiency and ease of implementation, remain an important approach for rapid risk screening over large areas. Physically based models can capture the underlying relationships among rainfall infiltration, pore-water pressure evolution, and slope failure, but their applications are constrained by the availability of model parameters and computational efficiency. Data-driven models show considerable potential for identifying complex nonlinear relationships and processing multi-source information; however, challenges remain regarding data quality, physical interpretability, and model generalization. In recent years, hybrid early warning models integrating physical mechanisms, artificial intelligence algorithms, and multi-source monitoring data have increasingly emerged as a research focus, providing new opportunities to improve the accuracy and reliability of regional-scale prediction of clustered landslides. Future research should further promote the deep integration of physical models and artificial intelligence, develop efficient three-dimensional dynamic simulation approaches, improve multi-source data assimilation and uncertainty quantification techniques, and advance early warning systems from risk identification toward mechanism-based analysis and intelligent decision support. These developments will provide a scientific basis for regional geological disaster risk management under extreme rainfall conditions.