A field experiment for calibrating landslide time-of-failure prediction functions

• Description of four years of continuous slope monitoring using an integrated platform.
• Investigation of the displacement of natural and anthropogenic slopes.
• Parametric analysis of forecasting methods for landslide time of failure failure prediction.
• Back-analysis of shallow small landslides and deformation of anchored bulkheads.
• Calibration of empirical parameters A and α for the Voight function.

Over the last decades, time-of-failure semi-empirical prediction functions have been developed and applied to different landslides with mixed results. In this study, a field experiment was carried out to calibrate these functions with the simultaneous consideration of small-size landslides and landslides that occur on slopes modified by human activities. Four years of continuous monitoring using an integrated platform consisting of both traditional sensors (i.e., inclinometers, piezometers, load cells, topographic measurement) and innovative remote-sensing equipment (i.e., Terrestrial SAR Interferometer) resulted in the collection of a notably large amount of data. Several landslides affecting different slopes (i.e., cut slopes, cut slopes covered by spritz-beton and slopes stabilised by anchored bulkheads) were observed as part of the experiment, thus facilitating the inference of detailed information for the pre-failure behaviour. Nine landslides were back-analysed, thus allowing for calibration of the failure prediction functions for different types of slopes. From these observations, it was found that events occurring on slopes modified by human interventions could be effectively predicted using the Voight function if suitable parameters are used. As a general remark, the landslides that originate from cut slopes in natural terrain behave similar to large landslides reported in the literature (similar values of A and α) while landslides that originate from cut slopes covered by spritz beton and slopes stabilised by anchored bulkheads show α values that are significantly lower and A values that are significantly higher than those of landslides on natural terrains.