GIS-Based Methods For Fractal Analysis And Hazard Estimation Of Regional Landslides

Landslide is a common geological disaster occurring regionally and frequently. It usually produces serious damages to factories, mines, cities and towns, villages, farmlands, transportation, communication, and even blocks rivers and destroys the water conservancy projects It occurred regionally and frequently in the mountainous areas of China. Therefore, it is of importance and significance to study the landslide so as to improve the landslide prevention and control, and to guarantee the engineering safety and the environmental protection.Landslide is be prone to occur and to cause natural hazards in the Three Gorges reservoir region, where there have varied topographies, complicated geological structures, and heavy annual rainfall. Especially since 1980's, the landslides occur more frequently to cause serious damages to the local communities. The landslide hazards often destroy houses, road, hydraulic supply lines, power-supply systems, bridges, docks, factories, mines, cities and towns, villages, farmlands, and vegetations. Some landslides swell damages to boats and villages nearby and even block up the rivers. In this way, the landslide hazards are the important issue to influence the settlement project, shipping and ecological environment in the Three Gorges reservoir region. It is thus important to study the landslide susceptibility assessment in the area to reduce the landslide hazards, and to ensure the reservoir safety.The landslide is a kind of complicated phenomenon with nonlinear inter-reaction. The traditional theories and methods are difficult to study the uncertainty characteristics of dynamic evolution of the landslide. Recently, the fractal theory has become a new method to study such complicated phenomenon as earthquake and landslide. However, there are still some key issues as below needed to further study.(1) The previous studies mainly focused on the fractal characteristics of geometric shape and spatial distribution of the landslide, but the knowledge in occurring conditions and inducing factors (landform, drainage network, and fault, etc.) of landslide are still unclear. Such knowledge is quite important to help to indicate the occurring conditions and evolution regulation to prevent and control the landslide hazards.(2) The previous studies are limited to simple description of the fractal characteristics of the landslide and have not obtained full understanding of the relationship between the fractal characteristics and the stability of the landslide. Their research results on the fractal characteristics of the landslide are unable to be applied to landslide susceptibility assessment effectively.The landslide susceptibility assessment is one of the effective approaches applied worldwide to prevent and reduce the landslide disasters, and it becomes one of the hot spots in research on landslide in recent years. But the occurrence of the landslide is controlled by lots of natural and man-made factors, and the relationship between the factors and the stability of the landslide is quite complicated. And the limitation of a single assessment method influences directly the reliability and accuracy. So the key technology and method to assess the landslide susceptibility effectively is the hot spot and difficulty in landslide study. There are still some issues as below needed to further study.(1) To determine the index system of landslide susceptibility assessment. It's quite important to identify and select the occurring factors of the landslide in landslide susceptibility assessment. But the previous methodology selects the assessing factors only with researcher's experience subjectively, without evaluating the assessing factors objectively and quantitatively, which makes the assessing results contain much more wrong conclusions. So it is a key issue needed to solve to determine the index system of landslide susceptibility assessment objectively and quantitatively.(2) To determine the weights of the assessing factors of landslide susceptibility assessment. It is also important for the assessment results to determine the weight of the assessing factors of the landslide in landslide susceptibility assessment. There are lots of quantitative and qualitative assessing factors, which are difficult to deal with the weights of the assessing factors. So it is a difficulty needed to solve to determine the weights of the assessing factors of landslide susceptibility assessment.The main study contents and conclusions in this dissertation are as follows:(1) Detailed studies are on the fractal characteristics of geometric shape and spatial distribution of the landslide, the occurring conditions and inducing factors as landform, drainage network and fault.(2) The relationship between the distribution of the landslides with the theory of the variable dimension fractal is studied.The study results indicate that the distribution of the landslides and slope, lithology and elevation and the cumulative distance from the river are in second-order cumulative total fractal distribution, and that the distribution of the landslides and aspect is in first-order cumulative total fractal distribution. The results show that slope, lithology and elevation and the cumulative distance from the river are more important to contribute to the occurrence of landslides than aspect, which provides the fractal basis to establish the index system of landslide susceptibility assessment.(3) The index system of landslide susceptibility assessment in the study area involves rock types and structure (lithology and fault), valley slope geometry and landform (aspect, slope, elevation and drainage network), hydrogeological conditions (Normal Different Vegetation Index, NDVI), dynamical factors (distance to river) and the fractal characteristics of the inducing factors (the fractal dimension of the small watershed).(4) Landslide susceptibility assessment based on weights of evidence modeling was involved.The weights of evidence model has been used for evaluating the contribution of the inducing factors to the occurrence of landslide to determine the assessing factors, and for landslide susceptibility assessment. According to the calculated weight values, the most important factors determining susceptibility are distance from river, elevation, lithology, distance from faults interface, drainage network and the fractal dimension of the small watershed. However, it is the combination of two or more classes of factors, which has the strongest influence on susceptibility and is not one single factor alone.The weights of evidence vary widely with different factors of evidence. The weights of evidence for lithology and for cumulative distance from faults interface are relatively very large, but they are smaller than those for cumulative distance from the Yangtze River. The weights of evidence for cumulative distance from the Yangtze River are the largest, which indicates that the river erosion is the most important factor controlling the occurrence of landslides in the study area, which is coincident with the actual fact that the landslides in the study area are mostly located along the banks of the Yangtze river. The calculated results show that the elevation is also an important factor controlling the occurrence of landslides in the study area. The calculated results show that the aspect and NDVI have certain influences on the occurrence of landslides, but the influence is relatively very small.There have differences in the weights of evidence among the different classes of the same factors of evidence. For example, the calculated results show that the weights of evidence for lithology of the Jialingjiang Formation (Tjs) of lower Triassic, the Badong Formation (Tb) of middle Triassic, the Tongzhuyuan Formation (Jt) of lower Jurassic, the Niejiashan Formation (Jn) of middle Jurassic and the 500m distance from interface between Permian and Carboniferous, Devonian and Silurian periods are much higher than any other lithology, which is coincident with the actual fact that the landslides are mostly located in the above lithology in the study area. The calculated results indicate that the river erosion becomes more and more violent and the slope becomes more and more unstable while the distance from the river is nearer and nearer, which is coincident with the occurrence of the already known landslides.According to the landslide hazards zonation mapping, there are three potentially dangerous areas with landslides in the study area. The first one is the region between Xietan and Huanglashi, where there exist such famous landslides as Huanglashi landslide, Fanjiaping landslide and Daping landslides. The second one is the region around Xiangxi town with the important Xiangxi landslides. And the third area is located in the region from Xintan to Lianzhiya with the famous Xintan landslide and Lianzhiya landslip. The study result is very coincident with the occurrence of the known landslides in the study area, which shows that the methodology is very useful and accurate and that the approach is appropriate for zonation of the landslide hazards.The study results are quite coincident with the occurrences of the known landslides, which show that the results are very accurate and have good quality. As a data-driven model, the weights of evidence model could help choose the main factors and calculate the weight values well and truly, avoiding man-made subjectivity. So the methodology is very useful and accurate and the approach is appropriate for zonation of the landslide hazards.(5) Logistic regression, fuzzy logic and fuzzy weights of evidence are used for landslide susceptibility assessment in the study area. The research results are very coincident with the occurrences of the known landslides.The main innovations of this thesis are as follows:(1) The fractal characteristics of geometric shape and spatial distribution of the landslide, the occurring conditions and inducing factors as landform, drainage network and fault are integratively studied.(2) The theory of the variable dimension fractal was used for evaluating the relationship between the inducing factors and the occurrence of the landslide to provide the fractal basis to establish the index system of landslide susceptibility assessment.(3) The index system of landslide susceptibility assessment in the study area involves rock types and structure (lithology, fault), valley slope geometry and landform(aspect, slope, elevation and drainage network), hydrogeological conditions(NDVI), dynamical factors(distance to river) and the fractal characteristics of the inducing factors(the fractal dimension of the small watershed).(4) The data-driven model (logistic regression and weights of evidence), the knowledge-driven model (fuzzy logic) and data-knowledge-driven model (fuzzy weights of evidence) were applied for landslide susceptibility assessment in the study area. The research results are very coincident with the occurrences of the known landslides.