Towards more reliable Debris Flow Rainfall ID Thresholds under Changing Climate Scenarios

Climate change is expected to increase the likelihood of hydro-geomorphic hazards in active tectonic areas, particularly debris flows. Early warning systems are considered one of the most effective and economical methods for mitigating debris-flow risk. However, current approaches still face challenges in providing accurate quantitative predictions and are subject to considerable uncertainty due to limited observational data. In this study, we develop a new framework for predicting rainfall thresholds for debris-flow initiation by combining numerical simulations with machine learning methods. A small catchment in the Italian Dolomites was selected as a test site to evaluate the efficiency of the framework in areas with limited historical records. Preliminary results suggest that the rainfall threshold can be represented by a piecewise function with an inflection point rather than by the commonly used power-law relationship. Our results suggest that, in the Dimai catchment, rainfall intensity is the dominant factor controlling debris flow initiation for the most rainfall events lasting longer than one hour. While sensitivity analyses indicate that infiltration capacity acts as a key control by regulating the partitioning between infiltration and runoff, thereby influencing the rainfall intensity required to trigger debris flow initiation. These findings provide insight into the hydrological processes governing debris flow initiation and demonstrate the potential of the proposed framework for improving threshold-based early warning systems under limited data conditions.