Temporal prediction in landslides - Understanding the Saito effect

Of all natural hazards, landslides offer the best potential for prediction of the time of an event, and indeed successful predictions of the time of large-scale slope collapse have now been made for a number of landslides. In general these predictions are based upon the use of the approach first proposed by Saito in which the inverse of landslide strain rate is plotted against time, and the time of final failure determined by fitting a straight line to the data, and extrapolating the time at which v(-1) = 0. In this paper, an investigation of Saito linearity is undertaken using laboratory testing. Using undisturbed samples of materials from the basal shear region of landslides, stress path testing has been undertaken. 38 mm diameter samples have been tested using the pore pressure reinflation (sometimes termed the field) stress path, in which the deviatoric and the total mean stresses are kept constant, whilst pore pressure is elevated. This triggers failure, during which the development of axial strain with time is monitored. The results clearly demonstrate that Saito linearity is the result of brittle deformation processes associated with the formation of the shear surface within the landslide mass. In landslides in which no shear surface forms, the linear trend is not observed. The data suggest that the state and rate dependent friction model, which has been implicated as a mechanism in some studies, is not the cause of this behavior. Our tests show that the linear phase of deformation in v(-1)-time space is associated with stress concentration during the late stage of shear surface formation. These results provide enhanced understanding of the processes of shear surface formation, and the circumstances under which Saito linearity can be observed.