- Landslides are a pervasive hazard that can result in substantial damage to properties and loss of life throughout the world. To understand the nature and scope of the hazard, landslide hazard mapping has been an area of intense research by identifying areas most susceptible to landslides in order to mitigate against these potential losses. Advanced GIS and remote sensing techniques are a fundamental component to both generate landslide inventories of previous landslides and identify landslide prone regions. A Digital Elevation Model (DEM) is one of the most critical data sources used in this GIS analysis to describe the topography. A DEM can be obtained from several remote sensing techniques, including satellite data and Light Detection and Ranging (LiDAR). While a DEM is commonly used for landslide hazard analysis, insufficient research has been completed on the influence of DEM source and
resolution on the quality of landslide hazard mapping, particularly for high resolution DEMs such as those obtained by LiDAR.
In addition to topography, multiple conditioning factors are often employed in landslide susceptibility mapping; however, the descriptive accuracy and contribution of the data representing these factors to the overall analysis is not fully understood or quantified. In many cases, the data available for these factors may be of insufficient quality, particularly at regional scales. These factors are often integrated into a wide assortment of analysis techniques, which can result in inconsistent mapping and hazard analysis.
To this end, the principal objectives of this study are to 1) evaluate the influence of DEM source and spatial resolution in landslide predictive mapping, 2) asses the predictive accuracy of landslide susceptibility mapping produced from fewer critical conditioning factors derived solely from LiDAR data, 3) compare six widely used and representative landslide susceptibility mapping techniques to evaluate their consistency, 4) create a seismically-induced landslide hazard map for landside-prone Western Oregon, and 5) develop automated tools to generate landslide susceptibility maps in a regional scale.
In this study, semi-qualitative, quantitative and hybrid mapping techniques were used to produce a series of landslide susceptibility maps using 10 m, 30 m and 50 m resolution datasets obtained from ASTER (Advance Space borne Thermal Emission
and Reflection Radiometer), NED (National Elevation Dataset) and LiDAR (Light Detection and Ranging). The results were validated against detailed landslide inventory maps highlighting scarps and deposits derived by geologic experts from LiDAR DEMs. The output map produced from the LiDAR 10 m DEM was identified as the optimum spatial resolution and showed higher predictive accuracy for landslide susceptibility mapping. Higher resolution DEMs from LIDAR data was also investigated; however, they were not significantly improved over the 10 m DEM.
Next, a series of landslide susceptibility maps were compared from six widely used statistical techniques using slope, slope roughness, elevation, terrain roughness, stream power index and compound topographic index derived from LiDAR DEM. The output maps were validated using both confusion matrix and area of curve methods. Statistically, the six output maps produced, showed accepTable prediction rate for landslide susceptibility. However, visual effects and limitations were noted that vary based on each technique. This study also showed that a single LiDAR DEM was capable of producing a satisfactory susceptibility map without additional data sources that may be difficult to obtain for large areas.
In western Oregon, landslides are widespread and account for major direct and indirect losses on a frequent basis. A variety of factors lead to these landslides, which makes them difficult to analyze at a regional scale where detailed information is not available. For this study, a seismically-induced landslide hazard map was created using a multivariate, ordinary least squares approach. Various data sources, including
combinations of topography (slope, aspect), lithology, vegetation indices (NDVI), mean annual precipitation, seismic sources (e.g., PGA, PGV, distance to nearest fault), and land use were rigorously evaluated to determine the relative contributions on each parameter on landslide potential in western Oregon. Results of the analysis showed that slope, PGA, PGV and precipitation were the strongest indicators of landslide susceptibility and other factors had minimal influence on the resulting map. An automated tool kit was a byproduct of this analysis which can be used to simply the hazard mapping process and selection of parameters to include in the analysis.