Comparison of the landslide susceptibility maps using two different sampling techniques with the frequency ratio (Fr) method

Abstract

In order to reduce the damages caused by the landslides, it is very important to predict the landslide occurrences and to determine the landslide susceptibility areas by the current methods in the literature. In this respect, it is aimed to produce landslide susceptibility maps of Ulus district of Bartın where landslides develop in the same lithology. The important point of the study is that Chebyshev theorem is tested for selected study area in this study and the susceptibility map produced by this method is compared with the landslide susceptibility map produced by using data set with pixel counting technique within the entire landslide body which is frequently used in the current literature. A total of 195 landslides were mapped in the study area and two different sampling strategies, Chebyshev’s theorem and landslide mass were used in the determination of landslide and non-landslide areas. In this study, landslide susceptibility analysis has been done for the study area by using topographic elevation, aspect, curvature and NDVI parameters. In the susceptibility analysis using both sampling strategies, Frequency Ratio (FO) method, which is frequently used in literature, was used and two different susceptibility maps were produced. The performance of the susceptibility maps was evaluated according to Area Under Curve method (ROC-AUC) and the AUC values were determined as 0.78 for Chebyshev theorem and 0.72 for the sampling technique according to the number of pixels in the entire landslide mass, respectively. According to these values, both susceptibility maps were acceptable and the performance of the susceptibility map produced by sampling with the Chebyshev theorem is relatively higher than the other sampling method. This result shows that Chebyshev method used in the study is an alternative method that can be used effectively in landslide susceptibility mapping studies and that the susceptibility map produced by this method has a successful prediction capacity.