Landslide model performance in a high resolution small-scale landscape

Research output: Contribution to journalArticleAcademicpeer-review

27 Citations (Scopus)


The frequency and severity of shallow landslides in New Zealand threatens life and property, both on- and off-site. The physically-based shallow landslide model LAPSUS-LS is tested for its performance in simulating shallow landslide locations induced by a high intensity rain event in a small-scale landscape. Furthermore, the effect of high resolution digital elevation models on the performance was tested. The performance of the model was optimised by calibrating different parameter values. A satisfactory result was achieved with a high resolution (1 m) DEM. Landslides, however, were generally predicted lower on the slope than mapped erosion scars. This discrepancy could be due to i) inaccuracies in the DEM or in other model input data such as soil strength properties; ii) relevant processes for this environmental context that are not included in the model; or iii) the limited validity of the infinite length assumption in the infinite slope stability model embedded in the LAPSUS-LS. The trade-off between a correct prediction of landslides versus stable cells becomes increasingly worse with coarser resolutions; and model performance decreases mainly due to altering slope characteristics. The optimal parameter combinations differ per resolution. In this environmental context the 1 m resolution topography resembles actual topography most closely and landslide locations are better distinguished from stable areas than for coarser resolutions. More gain in model performance could be achieved by adding landslide process complexities and parameter heterogeneity of the catchment.
Original languageEnglish
Pages (from-to)73-81
Publication statusPublished - 2013


  • physically-based model
  • soil redistribution
  • new-zealand
  • shallow landslides
  • sediment yield
  • dem resolution
  • stability
  • hazard
  • prediction


Dive into the research topics of 'Landslide model performance in a high resolution small-scale landscape'. Together they form a unique fingerprint.

Cite this