A predictive model of summer debris flows using machine learning techniques

Abstract The Alpine region of Aosta Valley has an early warning system to issue hydrogeological alerts up to 36 hours in advance based on the output of hydrological models and rainfall thresholds. However, those thresholds generally do not apply to summer debris flows, which can also be triggered by intense rainfalls of short duration, with cumulative precipitation lower than 20 mm. Therefore, it is necessary to formulate a specific predictive debris-flow model, which takes into account other possible triggering factors. In this study, we have developed a predictive model for debris flows with machine learning techniques, using a detailed dataset composed by a variety of geomorphological and hydro-meteorological variables. The variables of the dataset were collected from daily measured and modelled data for all of the 91 drainage basins in which at least one debris-flow event was generated during the time period considered in this study (2009–2019). The performance of the model, using different machine learning techniques, was evaluated and the most suitable model was chosen to be experimentally implemented in the existing early warning system of the region. The output of the model provides a debris-flow probability (DFP) for individual basins computed from the geomorphological and hydro-meteorological input variables.