UseCase » State of the Art

Landslide hazard constitutes the main technical challenge in landslide risk assessment. Since the 1970’s, various landslide hazard assessment models have been developed. They can be classified in different ways; one way is to classify them into expert-based, statistical and deterministic models:

In many countries (France, California, New-Zealand, Australia, Switzerland…), expert-based models have been widely used for hazard mapping (Van Westen, 2006). From the analyses of geomorphological parameters and recorded landslide data, different zones with their degrees of landslide susceptibility are defined through expert opinion. Relevant hazard assessments were made, relatively to available data and technological means. However, the qualitative and subjective aspect of these methods makes reproducing and exploiting them difficult.

Statistical models consist essentially in comparing a landslide inventory map with maps of possible terrain factors contributing to landslides, in order to find relations between the density of events and the parameters characterizing the factors. Progress in GIS (Geographical Information System) technology has enabled these methods to develop considerably. Statistical models are based on either bivariate or multivariate statistical analyses. New methods, such as logistic regression and artificial neural network classifiers, are developing. One problem with statistical approaches is that they can lead to an inappropriate consideration of physical processes (Dai, 2002). Moreover, it is difficult with statistical models to take account of the variations in the controlling conditions, for example to predict the evolution of hazard through time.

Deterministic models are based on mechanical slope stability analyses. The two main types of mechanical analyses are limit equilibrium analyses and those which simulate slope movement (Duncan, 2005). Deterministic analyses can be carried out within a probabilistic framework in order to take parameter uncertainties and variability into account, by using calculation methods such as FOSM, point estimate method, Monte Carlo simulation (Aleotti, 1998). A deterministic approach seems to be the most suitable approach to develop a methodology as objective as possible and relevant with respect to landslide physical mechanisms. However, deterministic models require a large amount of geomechanical input data, which are most of the time not totally available for technical or economic reasons.