RAINFALL INDUCED SHALLOW LANDSLIDE FORECASTING IN LARGE AREAS: APPLICATION OF THE TRIGRS MODEL OVER A BROAD AREA OF POST-OROGENIC QUATERNARY SEDIMENTS

This report describes an approach for calibrating a deterministic model to assess shallow landslide susceptibility over a 550 km² area of similar hydrogeologic properties. We used the Transient Rainfall Infiltration and Grid-based Regional Slope-stability (TRIGRS) model (Baum et al., 2002) to compute infiltration-driven changes in factor of safety, in a hilly-coastal portion of the Marche Region, central Italy. This area of the Esino River basin covers and is characterized by post-orogenic quaternary sediments prone to rainfall-induced shallow landslides. TRIGRS combines an infinite-slope stability analysis with a one-dimensional analytical solution for vertical infiltration under saturated or unsaturated soil conditions. We assumed saturated initial conditions and finite basal boundary depth. The input variables, such as mechanical, hydrological, and storm properties, have been collected together with a 20m DEM and a hydrogeological map of the study area. Published material properties data were obtained from neighboring areas, for soils that match the descriptions of the hydrogeologic map units, and were statistically analyzed and grouped into quartiles to set up a range of values to test. We ran the model for representative landslide-prone grid cells in each of the units for different rainfall scenarios. We compared modeled (pressure head, factor of safety) and observed (landslide occurrence) responses of each map unit to examine the model’s sensitivity to variations in material properties and to identify the most probable quartile. Using these values, we ran the model for the entire study area and fitted a soil depth model to the soil properties and landslide depth. The model’s output data were compared to a shallow landslides inventory. ROC curves for both the complete database and for particular rainfall events show that TRIGRS can be successfully used to build a susceptibility map and possibly predict rainfall-induced landslides over large regions even where geotechnical and hydraulic properties data are not available. Interpretation of the model’s output in a probabilistic framework accounts for uncertainties in the material properties and rainfall distribution as well as temporal changes in topography or subsurface conditions that are not represented in available geographical datasets.