“…First is the methodological incoherence in defining ID threshold values, e.g., as lower envelopes below which no landslides occur, upper envelopes above which landslides always occur, or intermediate curves defined by statistical methods. Examples range from simple visual fits [e.g., Caine , ] to methods based on Bayesian inference to define a minimum threshold [e.g., Guzzetti et al ., ; Brunetti et al ., ] and frequentist approaches which associate thresholds with different exceedance probabilities [e.g., Brunetti et al ., ; Meyer et al ., ; Peruccacci et al ., ; Brunetti et al ., ; Gariano et al ., ; Piciullo et al ., ]. Some studies are only based on triggering events, i.e., rainfall associated with landsliding, while others take into account also nontriggering events by Bayesian approaches [e.g., Berti et al ., ] or with statistical methods which maximize the probability of detection and minimize false alarms [e.g., Segoni et al ., ; Staley et al ., ; Corsini and Mulas , ].…”