| Abstract |
Vietnam is regarded as a country strongly impacted by climate change. Population and economic growth result in
additional pressures on the ecosystems in the region. In particular, changes in landuse and precipitation extremes
lead to a higher landslide susceptibility in the study area (approx. 12,400 km²), located in central Vietnam and
impacted by a tropical monsoon climate. Hence, this natural hazard is a serious problem in the study area. A prob-
ability assessment of landslides is therefore undertaken through the use of bivariate statistics. However, the land-
slide inventory based only on field campaigns does not cover the whole area. To avoid a systematic bias due to the
limited mapping area, the investigated regions are depicted as the viewshed in the calculations. On this basis, the
distribution of the landslides is evaluated in relation to the maps of 13 parameters, showing the strongest corre-
lation to distance to roads and precipitation increase. An additional weighting of the input parameters leads to
better results, since some parameters contribute more to landslides than others. The method developed in this
work is based on the validation of different parameter sets used within the statistical index method. It is called
“omit error” because always omitting another parameter leads to the weightings, which describe how strong
every single parameter improves or reduces the objective function. Furthermore, this approach is used to find
a better input parameter set by excluding some parameters. After this optimization, nine input parameters are
left, and they are weighted by the omit error method, providing the best susceptibility map with a success rate
of 92.9% and a prediction rate of 92.3%. This is an improvement of 4.4% and 4.2%, respectively, compared to the
basic statistical index method with the 13 input parameters. |
? |