Deterministic and Probabilistic Slope Stability Models Forecast Performance at ~1:5000-Scale

Deterministic methods are appropriate for analyzing specific slopes at site-scale where geotechnical parameters are better known. Probabilistic techniques provide better results than deterministic methods at regional scales (1:10,000–1:50,000). However, the performances of deterministic and probabilistic methods at large scales (e.g. 1:5000-scale) are not well-known. We applied GIS-based deterministic (WEDGEFAIL, SAFETYFACTOR, SHALSTAB) and probabilistic (Likelihood ratio) methods to a mountain road of 14 km in the Alpujarras region (S Spain) to investigate the behavior of these models at detailed scales. The studied road stretch was affected by 111 landslides (7–8 landslides/km) during the 2009–2010 winter in a period of high precipitation. These landslides cut off the road in several points and disconnected the central region of Alpujarras from the main transport infrastructures. We delimited a small study area with only 4 km2 restricted to the slopes that cross the road where we gathered as much data as possible. Our results show that deterministic methods have less prediction capability at ~1:5000-scale than probabilistic methods and it seems that the needed effort to improve their results is not worthwhile. However, it must take into account that probabilistic methods need an inventory and they could not have been applied before the analyzed landslide event. As our results indicate, the deterministic methods, such as the SHALSTAB model, are reliable tools to make an evaluation of the stability of cut slopes in a roadway at project-scale.