Debris flows (DFs) are classified as mixtures of sediments and water that rapidly route along steep channels. In mountainous regions, the impulsive nature of DFs combined with their high devastating power threatens human life and facilities (Jakob & Hungr, 2005) and has a high socioeconomic impact (Thiene et al., 2017). In this work, we refer to runoff-generated DFs, that is, those formed after the entrainment of debris material into abundant runoff descending from rocky cliffs (Coe et al., 2008; Hürlimann et al., 2014; Kean et al., 2013). Doubtless, high-intensity and sub-hourly rainfalls (15-45 minutes) are the predominant factors for the triggering and magnitude of runoff generated DFs. The prediction of such phenomena has great relevance for limiting damages and victims in the threatened areas and increasingly enters the context of social challenges. Many villages, tourist resorts, and linear infrastructures are built on DF fans or intersect DF channels. The expansion of human activities (new settlements, infrastructures, and touristic facilities, often in the extreme alpine environment) increases the vulnerability of society to these phenomena (Franceschinis et al., 2020). Moreover, the frequency of these phenomena is growing due to climate change (Stoffel et al., 2014); it leads to the increasing of both extreme rainfalls (Floris et al., 2010) and rock wall collapses that provide sediments even in locations where they never occurred before (Gatter et al., 2018; Bernard, Berti, et al., 2019). The large number of sites, which are potentially threatened, and the inherent inflexibility of technical countermeasures characterized by high costs make early warning systems the most cost-effective measure to mitigate the risk associated with DFs (Sättele et al., 2015). Because of the short times of DF downstream routing, the prediction of occurrences should occur as soon as possible. The real-time evaluation of triggering rainfalls permits to launch alerts in a time shorter than that given during DF downstream routing by sensors positioned along DF flow paths. The most common way to predict DF occurrences is the use of rainfall-based thresholds in early warning systems (