Landslide Susceptibility Modeling at Catchment Level for Improvement of the Landslide Early Warning System in Norway

摘要

In recent years many severe landslide events have occurred in Norway.These events are mainly shallow debris, soil slides and debris flows, which have caused significant damage.To prevent such damage an early warning system (EWS) for shallow slides and debris flows is being developed by the Norwegian Water Resources and Energy Directorate (NVE) based on hydro-meteorological thresholds.The experience in a testing phase shows that the EWS overestimates the warning level especially in some areas which are not prone to landslides.Thus, the combination of the index with information on landslide susceptibility may allow improving the performance of the EWS.In this study, two regions Northern and South-Eastern Norway are investigated.It was decided that modeling landslide susceptibility at the level of first order catchments would be appropriate.The landslide inventories for Northern and South-Eastern Norway contain 759 and 1419 landslides which occurred in 284 and 536 catchments, respectively.Landslide susceptibility modeling is carried out using Generalized Additive Models (GAM).Explaining variables are the Quaternary map, land cover, average yearly rainfall, various water runoff variables, and various derivatives from the 15m x 15m digital elevation model (DEM) like e.g.slope and aspect.Results of susceptibility modeling are well proofed by AUROC values for Northern Norway of 0.87 for the train sample and 0.84 for the test sample and for Eastern Norway of 0.88 and 0.86, respectively.In addition, the hydro-geomorphological plausibility proofed to be good.Combining the susceptibility map with the warning levels to establish new warning levels shows a clear improvement.Furthermore, modeling landslide susceptibility at catchment level suits the aim of improving the early warning system much better than e.g.a pixel-based approach.