Forecasting the failure of large rock slides is difficult because of nonlinear time dependency and seasonal effects, which affect the displacements. Starting from the accelerating creep theory proposed by Voight, a method is suggested to forecast slope failures and to assess alert velocity thresholds using monitoring data. The 20 Mm(3) Ruinon rock slide (Valfurva, Central Alps, Italy), susceptible to evolve into a rock avalanche, is studied. Three different evolutionary patterns of displacements have been recognized through the analysis of the monitoring data for a 5 year period. Data representing the surface-based large-scale behaviour of the rock mass were fitted by power-law curves, according to the "accelerating creep" model. Voight's equation has been expressed in terms of displacement and used to fit the data by nonlinear estimation techniques. Values for the controlling parameters (A, alpha, t(f)), representative of the mechanical behaviour of the rock mass approaching failure, have been determined both for single and multiple accelerating phases. "Characteristic velocity curves" have been computed by assuming these parameters are representative of the rock mass behaviour. Velocity threshold values for pre-alert, alert, and emergency phases have been computed. The method has been validated by collecting and analysing literature data for historical rock slope failures.
