Original paper

Assessing landslide vulnerability using bivariate statistical analysis and the frequency ratio model. Case study: Transylvanian Plain (Romania)

Roșian, Gheorghe; Csaba, Horváth; Kinga-Olga, Réti; Boţan, Cristian-Nicolae; Gavrilă, Ionela Georgiana

Zeitschrift für Geomorphologie, NF Volume 60 Issue 4 (2016), p. 359 - 371

published: Dec 1, 2016
manuscript accepted: Sep 22, 2016
manuscript received: Aug 1, 2016

DOI: 10.1127/zfg/2016/0404

BibTeX file

ArtNo. ESP022006004005, Price: 29.00 €

Download preview PDF Buy as PDF

Abstract

Abstract Landslides are among the most destructive natural hazards in several regions. Here we summarize our findings regarding this phenomenon in the Transylvanian Plain (Romania) using two susceptibility models: the statistical index and the frequency ratio model. Using Esri’s ArcGIS Raster Calculator tool we generated susceptibility maps by summarizing the following twelve landslide predisposition factors: lithology, soil type, fault distance, drainage network distance, roads distance, land use (Corrine Land Cover and NDVI), slope angle, aspect, elevation, plan curvature and soil erosion (RUSLE). The landslide susceptibility has been assessed by computing the values for each class of the predisposing factors and thus evaluating the distribution of the landslide zones within each factor, using Esri’s Tabulate Area Tool. The extracted predisposing factors maps have then been re-classified on the basis of the computed values in a raster format. Finally, the landslide susceptibility map has been reclassified into five classes using Natural Breaks (Jenks) classification method. The model performance was assessed with Receiver Operating Characteristic (ROC) curve and the R-index. The models with high number of factors had the lowest accuracy (AUC values being <0.8). The best frequency ratio model (AUC = 0.884) contained only three factors (slope, aspect, elevation) while in the case of the statistical index model the best model (AUC = 0.879) contained four factors (slope, aspect, elevation and NDVI). A significant part (33%) of the study area is characterized by a high to very high degree of susceptibility for landslides.

Keywords

statistical indexfrequency ratioGISlandslide susceptibilityROC curve