Development of Shallow Landslide Runout Distance Models Based on Climate Change Scenarios (Eocene Flysch Facies, Western Black Sea Region)

Abstract

The shallow landslides, increased by climate change, currently rank high in the list of natural disasters that humanity frequently struggles with, posing significant irreparable problems for many societies. The aim of the dissertation is to test empirically-statistically developed shallow landslides runout distance models to estimate possible damages by shallow landslides for large area. Eocene flysch facies, cover a quite extensive area and determined considering the geological natural boundary in Western Black Sea region of Türkiye, are exceedingly active with respect to shallow landslides occurrence. The shallow landslide inventory for the studied region was primarily prepared. Considering that the field covers a very large area, the field was divided into three sub-basins in order to model more realistically. In addition, descriptive statistical evaluations on the sub-basin basis regarding the travel angle, depth and observed runout distance of the shallow landslides in the prepared inventory were separately made for each three sub-basins. The testing of the runout distance probability assessment of shallow landslides requires to conduct the process of the detection of shallow landslides initiation points and their propagations. It is also necessary to generate a shallow landslide susceptibility map to allow for the most reliable prediction of possible initiation points of shallow landslides. Shallow landslides susceptibility maps were created using machine learning logistic regression method to determine the shallow landslide initiation points. This study was conducted based on two phases, one with Representative Concentration Pathway (RCP) scenario values and one without. Critical threshold values for shallow landslide initiations selected by considering shallow landslide susceptibility maps and RCP scenarios’ precipitation values. In the initial analysis, it was assumed that only the shallow landslide susceptibility is required to be greater than 0.70 in order for the cell to be accepted as the shallow landslide initiations. In the reckon with RCP scenarios, it was assumed that if the cell's shallow landslide susceptibility value is greater than 0.70 and it receives more than 81 mm of precipitation according to the RCP rainfall scenarios, that cell is accepted as the shallow landslide initiations. In the second stage of the study, runout distance empirical probability models were prepared for both models with RCP scenarios and without RCP scenarios. The Flow-R 1.0.0 software was utilized by applying Modified Holmgren algorithm and Simplified Friction-Limited Model (SFLM) algorithm parameters, aiming to offer an empirical estimation of runout distances during the propagation process. Two types of parameter sets were created, namely the debris flow parameters model and the shallow landslide parameters model for runout distance estimation, and their comparative evaluations were conducted. The determination coefficients for predicting the maximum runout distance probability of shallow landslides and debris flows using the empirical runout distance model were found to be 0.62 and 0.64, respectively. When the obtained results are also assessed by reckoning with RCP scenarios, it is manifested that the possible shallow landslides initiations and their runout distances will decline in the future according to RCP scenarios.