Research and Application of the Rainfall Intensity-Duration-Hazard Formative Environment (I-D-H) Early Warning Model for Clustered Landslides

[Objective] Rainfall-induced clustered landslides are characterized by small volumes, rapid onset, and large quantities. Traditional early warning models based on rainfall thresholds frequently neglect the variations in hazard-formative environments, leading to high rates of missed and false alarms, which underscores the urgent need to develop a more precise early warning model. [Methods] Based on historical data of rainfall-induced clustered landslides in Guangxi, a comprehensive hazard-formative environment factor (H) characterizing the susceptibility of geological-topographical conditions was extracted, and a three-dimensional early warning framework coupling rainfall intensity, duration, and the hazard-formative environment (I-D-H model) was proposed. On this basis, high-resolution environmental raster data of Beiliu City and recent landslide samples were integrated for localized parameter optimization to construct a refined I-D-H model perfectly tailored to the local hazard-formative background. A multi-scale application and validation was subsequently conducted utilizing the "June 9" extreme rainstorm event. [Results] At the regional scale, the proposed model accurately captures the differential impacts of the hazard-formative environment, improving the accuracy of yellow and higher-level warnings from 87% to 94% with zero missed alarms. At the site-specific scale, the model accurately identifies the red warning level for unstable slopes several hours in advance, effectively avoiding false alarms for adjacent stable slopes. Quantitatively, the Area Under the Curve (AUC) value reaches 0.829, demonstrating a predictive performance significantly superior to that of the conventional two-dimensional I-D model. [Conclusion] By dynamically adjusting critical rainfall thresholds, this model achieves a methodological leap in landslide early warning from regional macroscopic guidance to localized refined characterization, providing reliable technical support for the precise "site-to-regional" prevention and control of clustered landslide hazards.