Bivariate landslide susceptibility analysis in the Lorestan Arc (Zagros Mountains, Iran)
Keywords:Frequency Ratio, landslide susceptibility, bivariate analysis, steepness index, Zagros Mountains
Landslide susceptibility analysis based on the assessment of a quantitative relationship between multiple controlling factors and landslide occurrence is a consolidated approach for land-use planning in risk mitigation. The Zagros Mountain range (Iran) is one of the most spectacular examples of a landscape whose evolution has been controlled by the erosion of rock outcrops and the growth of thrust-fold structures (predisposing factors for gravity-driven deformations). This paper covers a preliminary study on the landslide susceptibility of the Lorestan region of the Zagros Mountains. Use was made of a bivariate Frequency Ratio, computed on a 30 m pixel size Shuttle Radar Topography Mission Digital Elevation Model. In particular, reliance was made on an unpublished inventory of landslides in the Lorestan Arc. Landslide susceptibility was assessed in the “Falls” (323) and “Slides” (297) categories, the numbers of which (unlike those of other categories) were regarded as suitable for robust modelling. A multi-parametric analysis was carried out to determine the susceptibility of each type of landslide to a set of contributing factors. These factors, most of which are commonly used in the literature, can be grouped in two main categories: a) static, including morphological, hydraulic, and geological factors, and b) pseudo-dynamic, including distance from active faults and steepness index (ksn), a morphometric tectonic uplift proxy computed along stream networks. The latter were considered as potentially landslide-triggering factors in the medium-long term, combining both the effects of transient tectonic forces of individual seismic events and morphotectonic effects on the slope-valley system. The statistical significance of these pseudo-dynamic factors was initially demonstrated via a univariate logistic regression with a randomly generated set of “stable” and “unstable” points. Then, models were built, and the importance of each conditioning factor was assessed by calculating the contribution of the factors in determining the mean landslide susceptibility index value in actual landslides, and the percentage of landslide cases in which the frequency ratio relative to each variable was above 1. With regard to the “Falls” category, the model showed (with a high reliability) that susceptible areas were those with a slope angle greater than 38°, associated with a 5000 m radius relief energy of about 1100-1550 m and a median fault distance of about 10 km. Conversely, the susceptibility model obtained for the “Slides” category showed that highly susceptible areas were those with slope angle values of 18°-35° and featuring outcrops of the most resistant lithotypes, such as limestone and marly limestone, especially close to flatiron landforms. The resulting model accuracy (validated with the prediction rate curve method on a sample dataset not used in the susceptibility analysis), was equal to 0.94 for “Falls” and 0.77 for “Slides”, indicating a good prediction accuracy.
How to Cite
Copyright (c) 2021 Italian journal of engineering geology and environment
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.