J. Geogr. Sci. 2012, 22(4): 617-629 DOI: 10.1007/s11442-012-0951-0 © 2012 Science Press Springer-Verlag Received: 2011-03-16 Accepted: 2012-02-20 Foundation: University of Firenze, Earth Science Department, Firenze, Italy Author: Rosi Ascanio (1982-), Ph.D. Candidate, specialized in landslide mapping and monitoring. E-mail: ascanio.rosi@unifi.it www.geogsci.com springerlink.com/content/1009-637X Statistical and environmental analyses for the definition of a regional rainfall threshold system for landslide triggering in Tuscany (Italy) ROSI Ascanio, SEGONI Samuele, CATANI Filippo, CASAGLI Nicola Department of Earth Science, University of Florence, Florence, Italy Abstract: The aim of this work is the determination of regional-scale rainfall thresholds for the triggering of landslides in the Tuscany Region (Italy). The critical rainfall events related to the occurrence of 593 past landslides were characterized in terms of duration (D) and intensity (I). I and D values were plotted in a log-log diagram and a lower boundary was clearly noticeable: it was interpreted as a threshold representing the rainfall conditions associated to landsliding. That was also confirmed by a comparison with many literature thresholds, but at the same time it was clear that a similar threshold would be affected by a too large approximation to be effectively used for a regional warning system. Therefore, further analyses were performed differentiating the events on the basis of seasonality, magnitude, location, land use and lithology. None of these criteria led to discriminate among all the events different groups to be characterized by a specific and more effective threshold. This outcome could be interpreted as the demonstration that at regional scale the best results are obtained by the simplest ap- proach, in our case an empirical black box model which accounts only for two rainfall pa- rameters (I and D). So a set of thresholds could be conveniently defined using a statistical approach: four thresholds corresponding to four severity levels were defined by means of the prediction interval technique and we developed a prototype warning system based on rainfall recordings or weather forecasts. Keywords: landslide; threshold; rainfall; black box models; prediction intervals 1 Introduction Landslides are a recurrent phenomenon that causes damages and casualties, therefore great efforts are being made to set up early warning systems able to forecast their occurrence. This aim can be achieved using two different kinds of models, white box and black box models. White box methods owe their name to their physically based approach in which the trig- gering mechanism is modeled on as many details as possible, trying to take into account every factor and every physical process involved in the occurrence of landslides. Such mod-