Modeling the Conditional Probability of the Occurrences of Future Landslides in a Study Area Characterized by Spatial Data

The most crucial but difficult task in the analysis of the risk due to landslide hazard is the estimation of the conditional probability of the occurrence of future landslides in a study area within a specific time period given the presence of spatial and geomorphologic features. This contribution explores a modeling procedure for estimating that conditional probability. The procedure proposed consists of two steps. The first step is to divide the study area into a number of “prediction” classes according to the hazard level for the likely occurrence of future landslides. “Favourability Functions” based on the spatial and geomorphological data in the study area were used for the sub-division. The number of the classes is dependent on the quantity and quality of the input data. Each class represents a level of hazard with respect to the future landslides. We term it the “hazard-mapping step”. For this step, several quantitative models have been developed and the strategy is to reconstruct the typical settings in which the future landslides are likely to occur. The second step is to empirically estimate the conditional probability in each prediction class given the spatial and geomorphologic data based on crossvalidation techniques. For the second step, termed the “probability estimation step” the basic strategy of the cross-validation is to construct the prediction classes in the first step using the occurrences of the landslides from the first time-period and then to compare the prediction classes with the distribution of the landslide occurrences from the later time period. The statistics obtained from the comparison provides the crucial quantitative measure to estimate the conditional probability. We illustrate the modeling procedure using a case study, La Baie, Quebec in Canada.