Data driven bivariate landslide susceptibility assessment using geographical information systems: a method and application to Asarsuyu catchment, Turkey

In the last decades, landslide hazard assessment has attracted many researchers' attention. A number of parameters are suggested to be responsible to quantitatively explain the mechanism of landslides; many of these parameters are very important and factual. However, some data types and models are site-specific and could not be applied to different locations. Furthermore, the data stored in continuous parameter maps are divided into a number of classes arbitrarily, depending on the vision of the expert. Basically, this division controls the result of bivariate analysis. Besides, the responsible portion of the parameter map controlling the mechanism is also weighted arbitrarily. Based on these two facts, the class boundaries put a prejudice on the produced susceptibility/hazard maps, which result in dependence on the knowledge of the user rather than being dependent on the data and the fact itself. The aim of this study is to refine the previously defined methods in a more data-dependent trend. To achieve this goal, two new concepts: seed cells and percentile maps are introduced. Seed cells are the zones that are considered to represent the best undisturbed morphological decision rules (conditions before landslide occurs) and would be achieved by adding a buffer zone to the crown and flank areas of the landslide. To quantitatively classify the input parameter maps, the data distributions of seed cells in the parameter maps are divided into a number of classes on the basis of their distribution's percentile break-points upon which the parameter maps are directly dependent on the seed cell distributions, hence to the data itself.     

A comparison of the GIS based landslide susceptibility assessment methods: multivariate versus bivariate

The purpose of this study is to evaluate and to compare the results of multivariate (logical regression) and bivariate (landslide susceptibility) methods in Geographical Information System (GIS) based landslide susceptibility assessment procedures. In order to achieve this goal the Asarsuyu catchment in NW Turkey was selected as a test zone because of its well-known landslide occurrences interfering with the E-5 highway mountain pass.Two methods were applied to the test zone and two separate susceptibility maps were produced. Following this a two-fold comparison scheme was implemented. Both methods were compared by the Seed Cell Area Indexes (SCAI) and by the spatial locations of the resultant susceptibility pixels.It was found that both of the methods converge in 80% of the area; however, the weighting algorithm in the bivariate technique (landslide susceptibility method) had some severe deficiencies, as the resultant hazard classes in overweighed areas did not converge with the factual landslide inventory map. The result of the multivariate technique (logical regression) was more sensitive to the different local features of the test zone and it resulted in more accurate and homogeneous susceptibility maps.     

Evaluation of environmental parameters in logistic regression models for landslide susceptibility mapping

Abstract

The aim of this study was to determine how well the landslide susceptibility parameters, obtained by data-dependent statistical models, matched with the parameters used in the literature. In order to achieve this goal, 20 different environmental parameters were mapped in a well-studied landslide-prone area, the Asarsuyu catchment in northwest Turkey. A total of 4400 seed cells were generated from 47 different landslides and merged with different attributes of 20 different environmental causative variables into a database. In order to run a series of logistic regression models, different random landslide-free sample sets were produced and combined with seed cells. Different susceptibility maps were created with an average success rate of nearly 80%. The coherence among the models showed spatial correlations greater than 90%. Models converged in the parameter selection peculiarly, in that the same nine of 20 were chosen by different logistic regression models. Among these nine parameters, lithology, geological structure (distance/density), landcover-landuse, and slope angle were common parameters selected by both the regression models and literature. Accuracy assessment of the logistic models was assessed by absolute methods. All models were field checked with the landslides resulting from the 12 November 1999, Kayna?li Earthquake (Ms = 7.2).