Multi-Hazard Susceptibility Map Production for Urban Planning Purposes: Ankara Mamak Case

Abstract

Multiple hazards may affect urban areas and integration of susceptibility maps about natural hazards are required for proper site selection and urban planning. Likewise, geological and geotechnical parameters, development expenditures and the spatial pattern of existent social and technical infrastructure should be considered for this aim. Up-to-date land use land cover (LULC) maps and the natural hazard susceptibility maps based on this data can be acquired from high-resolution satellite sensors. In this thesis, landslide and flood hazard potential are considered for combined hazard susceptibility assessment to be used for an urban settlement, Mamak, which is currently developing for disaster management purposes. The flood susceptibility map of Ankara City was generated in a prior research with the Modified Analytical Hierarchical Process (M-AHP) method. The landslide susceptibility map was generated with the logistic regression method in this thesis. LULC map was obtained from Sentinel-2 images with the Random Forest method as an image classification technique. The first and second topographic derivatives acquired by a digital terrain model (DTM) and lithology information produced by the General Directorate of Mineral Research and Explorations were employed for the generation of landslide susceptibility maps. A combined hazard susceptibility assessment was performed using the Mamdani Fuzzy algorithm and the outcomes were examined in the context of urban planning. According to the results, multi-hazard susceptibility assessment maps, which can be a base for city planning, can be created by combining a number of machine learning and expert-based learning methods.