LANDSLIDE MICROZONATION IN HIMALAYAN REGION USING GEOSPATIAL TOOLS

Abstract

Hilly regions are highly prone to natural disasters. Natural disasters like floods, forest fires, slope failures, erosion, and landslides are common phenomena in hilly terrains. Landslide is a disastrous phenomenon that is responsible for economic losses and loss of lives. Landslides are responsible for huge economic loss which makes them 3rd largest natural disaster after floods and earthquakes. The planners and engineers require information about the possibility of the occurrence of landslides in working regions. So, the landslide hazards must be planned carefully to avoid losses. Microzonation of landslides or slopes which are susceptible to failure is an important task for understanding and planning the mitigation measures for landslides. Microzonation of landslides is the identification of the potential occurrence of landslides in different areas. The microzonation maps of landslides represent the landslide susceptibility and distribution of previously occurred landslides. The main objective of the thesis is to implement and compare the mixed methods and quantitative techniques of landslide susceptibility mapping. Due to the high subjectivity of the opinion of the experts, the qualitative techniques give lesser accuracy as compared to the statistical method. So, it is also attempted to propose a hybrid technique for enhancing the accuracy of expert-based methods.In this thesis, causative factors of landslides are identified using historical landslide data. The causative factors of landslides considered in this study are slope gradient, slope aspect, relative relief, topographic wetness index (TWI), lithology, drainage density, proximity to the road, proximity to faults/lineament and land use of the study area. The causative factors of the landslides are divided into simpler sub-categories. For example, the slope is divided into sub-categories like 0º to 15º, 15º to 30º, etc. Slope, aspect, relative relief and TWI are extracted by processing the vi CARTOSAT DEM. The parameters that are not extracted from the digital elevation model (DEM) are converted into the digital format using the geographic information system (GIS). The impact of these causative factors on the occurrence of landslides is evaluated using an expert-based approach and mathematical approach. In this study, four models are implemented for the microzonation of landslides in the study area. A landslide inventory containing more than 1500 landslide events is prepared using previous literature, news reports, Geological Survey of India (GSI) practical sheets, google imagery and field survey. Analytic hierarchy process (AHP), frequency ratio (FR) and Shannon’s entropy models are used for landslide susceptibility mapping and a new technique by hybridization of Shannon’s entropy and AHP model is proposed taking a case study of Shimla region in Himachal Pradesh (H.P). However, the analytic hierarchy process (AHP) is a semi-qualitative model and an improvement over expert-based techniques, the rest three models are mathematical models. The weightage of causative factors and sub-factors are determined based on expert opinion and are checked for consistency in AHP. The weightage of causative factors and sub-factors in the other three models is obtained using mathematical relationships. Four landslide susceptibility maps for the study area are prepared and the performance of each method is evaluated using the receiver operation characteristics (ROC) curve. It can be observed that the frequency ratio (FR) model is the most effective approach in predicting the landslide susceptibility while the analytic hierarchy process (AHP) remained the least productive. The hybrid model i.e. SE-AHP model performed better as compared to the analytic hierarchy process (AHP) model. Shannon’s entropy model assigns weightage to the causative vii factors and sub-factors both, but still, the model’s accuracy is lesser compared to the frequency ratio (FR) model. The mathematical models require a well-distributed landslide inventory while it is not essential for the expert-based models. It is observed that the accuracy of the results in mathematical methods depends upon the distribution and accuracy of the landslide inventory while the accuracy depends upon the expert’s judgment in the case of expert-based methods. Some work has been reported related to mathematical and expert-based models but Shannon’s entropy has been used very rarely. The performance of the AHP model is improved significantly by the hybridization of AHP with Shannon’s entropy. The results of the study revealed that realistic weightage can be obtained only from an accurate and well-distributed inventory. Finally, the thesis presents a comparison of expert-based methods and mathematical methods for landslide susceptibility mapping. The study helps in identifying the contribution of causative factors in the occurrence of landslides. The output of the study helps in the demarcation of the zones of high landslide potential. This study also provides information that can be used by the researchers in understanding and choosing the suitable method for landslide susceptibility mapping. The newly proposed mixed technique in this study can reduce the subjectivity in the expert-based methods and improve the accuracy of the AHP model. The results of the study will also help the planners and risk managers for understanding the landslide potential in the study area.