Slope-specific rainfall thresholds for regional landslide early warning systems

Abstract

The increasing frequency and intensity of extreme rainfall events have amplified the demand for effective landslide early warning systems. However, traditional regional models often neglect the spatial variability of rainfall thresholds, resulting in reduced warning efficiency. This study proposes a region-based threshold calculation framework that incorporates slope-specific geoenvironmental characteristics and historical rainfall conditions, aiming to bridge the gap between regional-scale and slope-scale early warning approaches. The method first establishes base rainfall thresholds and quantifies the spatial variability of geoenvironmental factors (SVGF) using a widely adopted regional thresholding technique and the information value method for landslide susceptibility assessment. It then constructs the spatial variability of historical rainfall (SVHR) based on the maximum historical effective cumulative rainfall. Finally, slope-specific thresholds are derived by adjusting the base curves using variability coefficients obtained from SVGF and SVHR. The proposed framework was validated in six rainfall-induced landslide-prone counties in northeastern Chongqing, China. Results show that the method outperformed existing models in both the modeling dataset (accuracy = 79.07 %) and the prediction dataset (accuracy = 75.39 %). During extreme rainfall events, the average hit rate improved by 46.10 %, and the maximum AUC reached 0.9282—surpassing all other models. By extending traditional threshold frameworks to support slope-specific adaptation, the proposed method effectively integrates with existing thresholding and susceptibility models. It offers a technically sound and adaptable solution for landslide early warning, with considerable promise for practical application.