Assessment of landslide hazard in the environmental hotspot areas of the Kyrgyz Tien-Shan: Spatial analysis and Numeric modelling

This PhD research was initiated in parallel with the Nato Science for Peace project “Prevention of Landslide Dam Disasters in the Tien Shan, Kyrgyz Republic”. During this project a large amount of thematic data on slope instabilities and landslide dams in the Kyrgyz Tien-Shan was collected. The thesis uses this information to investigate applied and fundamental aspects of the regional mapping of the landslide susceptibility, hazard and, partially, risk. The target areas selected for the PhD research are the Mailuu-Suu River Valley (southern Kyrgyzstan) and the surroundings of the Minkush settlement (central Kyrgyzstan). They represent the areas of former uranium mining hosting numerous storage sites of tailing material and waste rock. Additionally, these areas are characterised by a high level of landslide activity induced by the human and environmental impacts. The landslides in those areas present a high risk to the society and infrastructure, potentially triggering serious environmental consequences. Structurally, the thesis is composed of two parts: Part A and Part B. These parts are different both with respect to the research methodology and the scientific targets. The common goal of both parts is the development effective techniques to regionally assess landslide hazard in mountainous areas marked by complex geomorphological, tectonic and geological settings. Due to the remote access, such areas are often lacking background information and in-situ data that are required for the precise prediction of landslide occurrence and impacts. In such cases the spatial analysis often helps to better characterise regional and, partly, local landslide susceptibility. The regional studies allow us to outline the basic relationships between affecting factors and landslide occurrence. Such results also become the basis for effective planning of in-situ investigations and localised measurements. Part A of thesis is focused on the applied aspects of spatial analysis. This part assembles a set of the well-established approaches and methods to model landslide susceptibility, hazard and risk. The studies start with the compilation of database spatially combining various thematic information. The first type of such thematic data is represented by the multi-temporal landslide inventories. These inventories are extracted applying different sources of remote sensing data, including the satellite and aerial imagery. The second type of thematic data describes the spatial distribution of factors affecting the level of landslide activity. The mapping of affecting factors is performed by applying conventional remote sensing techniques and spatial analysis tools. The collected thematic data are then used to extract the landslide susceptibility, hazard and risk maps. In total, four conceptually different approaches are applied to map the landslide susceptibility, both on a qualitative and quantitative basis. The results of the quantitative susceptibility mapping and thematic data are further used to calculate the landslide hazard for each part of the studied area. The calculated landslide hazard is characterized both by spatial and temporal components. The results of the landslide hazard assessment are finally used to estimate the risk of the direct impacts of landslides on selected exposed elements, including the uranium tailings sites. One of the well-established approaches used in Part A is the Newmark method. It maps the seismically triggered landslide susceptibility based on the computed co-seismic displacements. The simplicity of this method is attractive to many researches around the world studying seismically-triggered landslide hazards. Nevertheless, the simplifications adopted in this method strongly limit the reliability of the final predictions. In the second part of thesis we provide a critical overview of the Newmark method and attempt to propose conceptual improvements of the existing mapping practice. To reach the targeted challenges we combine the spatial analysis with the dynamic simulations in the 2D and 3D domains. The studied models represent the actual topographic and geologic settings of the landslide-prone slopes. The simulations provide acceleration time histories that are recorded in different parts of the model surface. The analyses of these records allow us to outline the amplification impacts related to the topographic and geological site effects. The modelled amplification factors are analysed with respect to their link to the local geological and topographic settings. Thus, we study how the structural or material settings of the model can impact the recorded geological amplification. Such impacting parameters can be represented by the layer thickness, inclination of the underground contacts or the contrast of the shear wave velocity (Vs) values. The 2D topographic modelling investigates the way how the pure topographic amplification factors can be predicted based on the surface morphology. The studies show that the surface curvature can be considered as a key parameter to predict the amplification factors. This idea is further investigated in the 3D modelling studies applying different input signals and the materials of varying Vs. The results of this modelling are integrated into a single database which is subjected to spatial analysis. This finally allows us to develop a simple proxy which maps the impacts of the topographic site effects based on morphological parameters extracted from the Digital Elevation Model of the target area. Additionally, the subset of the 2D dynamic tests analyses the shear displacements triggered by the seismic shaking. Those displacements are recorded in the models with simplified geometry, as well as in the ones presenting the real topographic settings. The recorded displacements are cross-correlated with parameters characterising the seismic impacts. The results of analyses allow us to develop a new law which can easily be used in the GIS-based studies. The proposed law applies a set of the conventional parameters as well as a novel predictor which has never been used by any regional law before.The conceptual improvements proposed in Part B are finally tested by applying them to the geodatabase compiled in the first part of thesis. Related results are then compared with those that were produced by using the conventional Newmark method. The validation tests should assess performance of the developed proxies to improve the predictions of seismically-triggered landslides