A reliable forecast of the failure stage of large rockslides is difficult, because of non-linear time dependency of displacements and seasonal effects. Aim of this paper is to suggest a practical method to prepare alert thresholds for large rockslides, assessing critical values of velocity for carrying out civil protection actions using monitoring data. Adopted data concern the 20 Mm(3) Ruinon rockslide (Valfurva, Central Alps, Italy), still evolving and suitable to originate a fast moving rock avalanche. Multitemporal analysis of aerial photos, LIDAR-ALTM laser topography, field survey and geomechanical analyses allowed to infer the rockslide kinematics and better understand data provided by a monitoring network including distometers, extensometers, GPS benchmarks and inclinometers. The analysis of displacement and rainfall data over five years (1997-2001) allowed to recognise three different evolutionary patterns of displacements, showing a continuously increasing rate since 1997. Data representing large-scale behaviour of the rock mass were fitted by power-law curves, according to the "accelerating creep" model by Voight, in order to evaluate a suitable failure time. This was hampered by the large seasonal deviations, which can significantly delay the occurrence of failure. Data were fitted using the Voight's equation, expressed in terms of displacement, through non-linear estimation techniques, in order to find values of the controlling parameters (A, a and If) suitable to represent the mechanical behaviour of the rock mass approaching the failure. This allowed to compute velocity-time theoretical curves and to define different velocity threshold values (pre-alert, alert and emergency) to be used for emergency management. (C) 2002 Elsevier Science Ltd. All rights reserved.
