LANDSLIDE SUSCEPTIBILITY EVALUATION IN A PART OF GARHWAL HIMALAYAS USING GIS

Abstract

Landslide susceptibility assessment is done based on three different approaches viz., Physical modelling, Statistical methods and Geomechanical methods. This is achieved through integration and manipulation of data using Geographic Information System (GIS) supplemented by other modes of computing and execution. A test area of 84km is selected in Garhwal Himalayan region covering two catchments of river Alkananda. Physical modelling involved two simulation programs in C++ programming language to determine rockfall velocity and runout zone estimation respectively. GIS is extensively used in data generation, visualization and interpretation for both the models. Simulation programs are formulated applying different conditions on the basic principle of potential influence of gravitational pull on a falling rock mass at a point, with respect to a point at relatively higher position. The first model estimated rockfall velocity considering all points of higher elevation as source points. The second model uses aspect based slope unit criteria to calculate rockfall velocity as well as estimate the runout zone. The first model estimated velocity in the range of 0 ms" at highest point tol80 ms"' in the valley floor. Abrupt change in velocity gradient is evident in isolated patches in the catchment owing to friction dependent variable incorporated in the model. The second model is treated separately for existing and potential source points and found to efficiently evaluate the physical process of rockfall in the catchment. The second approach is based on statistical methods. It is attempted through four different techniques utilizing GIS potential for raster based analysis. The first two methods namely, Information Value (Info Val) and Frequency Ratio (FR) are pixel based analysis involving statistics of different ratios of landslide pixel in parameter class, pixel count of each class domain and frequency of each class in the whole area. The third method, Certainty Factor (CF) is also based on landslide and class frequencies for estimating the probability of landslide occurrence in each pixel. Individual class CF's are combined based on certain rules to generate a susceptibility map by zoning the area under six predefined categories. The fourth method, Logistic Regression (LR) is a multivariate statistical technique which incorporates landslide densities under each class of parameters in a logit transformation equation to estimate the range of probability of landslide occurrence at each pixel. The first, second and fourth methods give a range of values in the output and hence is categorized into preferred susceptibility classes. The output of the third method falls in any of the six predefined susceptibility categories estimated at each pixel. The first two highest susceptibility classes in all the four analyses invariably account for the high count of landslide pixels irrespective of techniques. These four different methods are compared and validated. Out of the four methods, the highest value of confidence is found in the use of Certainty Factor. The third and final approach is based on GIS based assessment of geomechanical properties of rock slopes. It is addressed through analysis of two methodologies. The first method is aimed at analysis of slope failures in a small catchment in relation to the dip and azimuth of discontinuity and topographic slope. Dip and azimuth of discontinuity set is analysed with that of topographic slope to identify pixels falling in identical ranges. The output is a presentation of potential failure surface which is compared with the existing landslide map through proximity analysis. A very good relation could be established between the existing slope failures and the potential failure points. The second technique, Slope Mass Rating (SMR) based Geomechanical modelling is built on empirically established ratings of various categories in different parameters of rock mass classification. The ratings are incorporated in GIS through raster based analysis for slope stability assessment along a road section in the catchment. The resulting map is categorized by dividing the SMR values equally in to five classes. A good correlation is found between the failed slope faces and the low SMR values.