Production of Highway Landslide Susceptibility Map with Machine Learning Techniques: A Local Study from Türkiye, Artvin-Ardanuç Road Line

Landslide (landslide) is a natural event that occurs when the upper layer of the soil slips away when certain parameters are met. This natural event occurs in many places in the world. In Turkey, landslides are observed especially in the Eastern Black Sea Region. Therefore, a landslide susceptibility map was tried to be produced in order to investigate the question of how sensitive a piece of land can be to landslides as a region. In particular, it was tried to investigate how important a landslide susceptibility map can be in determining a highway line. In our study, the taxonomy of the 35 km road line between the Ardanuç District of Artvin Province, 65.36 km2 soil region area was determined by considering 11 elements such as altitude, aspect, soil moisture index, precipitation, curvature, curvature angle, land cover, lithology, distance to drainage networks, distance to fault lines, and slope. The landslide susceptibility maps produced were divided into five susceptibility classes as very high, high, medium, low and very low. The predictive skills of the susceptibility models were examined by supervised algorithms of machine learning such as linear regression, logistic regression, support vector machine, decision tree and random forest and XG Boost (extreme gradient boosting) which would be the most suitable model.