Comparative assessment of landslide susceptibility. Case study: the Niraj river basin (Transylvania depression, Romania)

Abstract

This study represents a comparison between two independent models used to evaluate landslide susceptibility in Romania: first, the model derived from the Romanian Governmental Decision no. 447/2003 (H.G. 447) and second, the bivariate statistical analysis. Considering the numerous objections to the first approach, which is also imposed by law, the accuracy of the results was analyzed using an alternative method which takes into consideration the reality from the field to a greater extent (the inventory of the existing landslides). The case study is focused on the Niraj catchment area (658 km2), a representative area for frequent landslide occurrence. The H.G. 447 model implies the estimation of the importance of eight factors involved in landslide occurrence: lithology, geomorphology, structure, hydro-climatic factors, hydrogeology, seismicity, forest cover and the anthropogenic factor. A thematic map was generated and analyzed for each one of the eight factors influencing slope instability and a specific coefficient was assigned. The statistical model, based on the bivariate probability analysis, was applied in order to predict the spatial distribution of the susceptibility classes. The probability of landslide occurrence was estimated based on the assumption that the prediction of the spatial distributions of landslides starts from the existing ones. In order to validate the model, the resulting maps were compared with the existing landslide maps: the relative landslide density index (R) and the relative operation curve (ROC) value were calculated, which indicate that the statistical model emphasizes a better correlation between the susceptibility classes and the active landslides (ROC value 0.972), the causative factors selected being relevant for the applied models.