Influence of DEM and Soil Property Uncertainty on an Infinite Slope Stability Model

M. Ulmer1, P. Molnar2 and R. Purves3
1 WSL Institute for Snow and Avalanche Research SLF
Flueelastrasse 11
7260 Davos Platz
Switzerland
Telephone: +41 81 417 0243
Fax: +41 81 417 0110
Email: ulmer@slf.ch

2 Institute of Environmental Engineering, Swiss Federal Institute of Technology Zurich
Wolfgang-Pauli-Str. 15
8093 Zurich
Switzerland
Telephone: +41 44 633 2958
Email: peter.molnar@ifu.baug.ethz.ch
3 Department of Geography, University of Zurich
Winterthurerstrasse 190
8057 Zurich
Switzerland
Telephone: +41 44 635 6531
Fax: +41 44 635 6848
Email: ross.purves@geo.uzh.ch

Landslides pose a worldwide threat to humans, infrastructure and agricultural land, with around 1800 people dying in landslides annually (Alexander, 1995). GIS-based slope stability models are widely used to identify areas prone to landsliding, however the predictions of these models, and indeed any models, are susceptible to a range of uncertainties, including that resulting from uncertainty or error in input data and error propagation within models (Heuvelink, 1998). Furthermore, model evaluation typically explores either a single set of parameters, or extremes in individual parameters through sensitivity analysis (e.g. soil parameters (Guzzetti, 2005)), but neglects a holistic characterisation of the nature and form of uncertainty of parameters
on model results. In this paper, we illustrate the importance of digital elevation model (DEM) uncertainty in landslide modelling, and compare it with the uncertainty in other model parameters. We also explore the relation between DEM resolution and prediction performance and methods to evaluate model results using ground truth. The analysis is based on a standard slope stability model using the 'infinite slope approximation', similar to SINMAP (Pack et al., 2005). The model produces a spatially distributed factor of safety (FS) which is the ratio of stabilizing and destabilizing forces on a
hillslope. The model is best suited to shallow rainfall-triggered landslides and in our study incorporates two parameters derived from a DEM: slope gradient and the topographic wetness index. Soil properties are represented by four parameters: soil thickness, cohesion, hydraulic conductivity and friction angle. The analysis focuses on a research area in the region of Napf, Switzerland, where a storm in July 2002 led to widespread landsliding (Rickli and Bucher, 2002).

AttachmentSizeHitsLast download
ulmer2009geomorphometry.pdf1.09 MB8601 week 2 days ago
3.333335
Your rating: None Average: 3.3 (3 votes)