Identification and characterization of rainfall events responsible for triggering of debris flows and shallow landslides

The aim of this study is the development of objective and replicable methodologies for the identification, analysis and characterization of rainfall events responsible for the triggering of shallow landslides and debris flows, in order to define empirical rainfall thresholds. The study area is the province of Trento (6208 km2), located in the north-eastern Alps, and characterized by complex orography, with 70% of the area at an altitude above 1000 m. A rigorous statistical methodology has been defined for the identification of the beginning of the triggering event, based on the critical duration, i.e. the minimum dry period duration separating two stochastically independent rainy periods. The critical duration has been calculated for each rain gauge of the studied area and its variability during the months of the year has been analyzed. An analysis of the rainfall spatial variability in a neighborhood of the landslide detachment zone has been carried out. The adopted methods are: the examination of the Monte Macaion radar maps during some summer convective events, the comparison of rainfall records of rain gauges located in a 10 km buffer around the landslide, and the calculation of the Pearson's correlation coefficient between pairs of neighboring rain gauges. The following rainfall thresholds have been then calibrated with the frequentist approach and compared: average intensity–event duration (I–D), which represents the rainfall event in its entirety, and intensity–duration associated with the event maximum return period (IRP–DRP), which considers the most critical portion of the event. In the absence of information about the landslide time of activation, the end of the triggering event has been identified using two criteria: the rainfall peak intensity and the last registration of the day. The methodology adopted for the objective identification of the beginning of the triggering event has demonstrated good applicability for rainfall induced landslides. During convective summer events the low representativeness of the rainfall information recorded at the nearest rain gauge with respect to the precipitation over the landslide source area has been evaluated as a critical issue for rainfall threshold definition.