Assessment of landslide processes on the territory of Tomsk by using GIS technologies

Abstract

Introduction: Currently, the city of Tomsk is extending its area of development, new housing blocks are appearing on previously undeveloped land (e.g. on the left bank of the river Tom), the central part of the city is being redeveloped and reconstructed. Without taking into consideration the dynamics, mechanisms, factors, and patterns for the development of dangerous natural and technogenic-natural processes and the forecasting of their development it is not possible to develop the territory qualitatively. The goal of the work is to establish patterns for the occurrence of landslides, to estimate the intensity of their development, and to predict the probability of their occurrence within the new boundaries of the city. Research methods: The assessment and forecast of landslide development were performed with the help of GIS technologies, which are an important tool for the city's management due to their ability to process and analyse multidimensional data regarding the geological environment. Two methods were used to determine the “weight” of factors contributing to the development of landslides: the analytical hierarchical process (AHP) and the frequency ratio (FR). The analysis of curves for Receiver Operating Characteristics (ROC) was used to validate the susceptibility maps. Results and discussion: A map of the distribution of landslides within the territory of the city was created. The map includes 25 landslide locations. These landslides were randomly divided into two parts: 17 landslides (70%) to train the model and 8 landslide locations (30%) to validate the model. Seven factors were selected to analyse spatial patterns that determine the development of landslides: slope angle, slope aspect, curvature, elevation, geological structure of the territory, types of filtration sections, and the distance to the river. A training set of data was used to create susceptibility prediction maps by ranking the factors by two methods. Spatial correlation between the location of landslides and contributing factors was identified using GIS-based statistical models. The quality of the models was evaluated by means of a ROC-curve analysis. The area under the curve was 0.750 for the AHP-model and 0.844 for the FR-model, which shows the high quality of the prediction maps. Thus, both methods have proved useful in assessing the susceptibility of areas to landslides. Conclusions: Landslide susceptibility maps for the territory within the new boundaries of the city were created for the first time and the quality of the models was checked. The maps are recommended to be used to monitor and manage swiftly the geological environment, to evaluate land in the urban area, and to teach the discipline “Engineering Geodynamics”.