Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.authorSaha, Ashis Kumar-
dc.guideCsanlovics, E.-
dc.guideArora, M. K.-
dc.guideGupta, R. P.-
dc.description.abstractHimalayas constitute the loftiest and the youngest mountain system on the Earth. Owing to their spread, elevation and general ruggedness, most of the regions in the Himalayas are not easily accessible, roads providing the only way of transport. Occurrence of landslides is one of the major problems in this region causing extensive damage to life, property and communication every year. It has been observed that most of the roads have been constructed disregarding the distribution of Landslide Hazard Zones in the region. Many sections of the roads get closed due to landslides in rainy season, thereby disconnecting many villages and towns in the area. Therefore, there is an immense need to develop a strategy utilising advanced techniques for route planning in landslide susceptible terrains, such as the Himalayas. Remote Sensing and Geographic Information System (GIS) techniques, by virtue of their numerous advantages, appear to be an automatic choice to tackle this problem, which requires efficient processing, interpretation and analysis of a large amount of spatial data from a variety of sources. The main objective of this research is to explore the potential of the advanced technologies - remote sensing and GIS, and to devise an automatic and intelligent approach for route planning in a hilly region prone to landslide hazards. The study area covers a small region of about 550 km2 in the Himalayas (Latitude 30°20'-30°34'N and Longitude 79°05'-79°22'E). It is a part of Chamoli and Rudraprayag districts of the newly formed state - Uttaranchal - of India. The terrain is highly rugged with elevations ranging from about 920 m to 4250 m above mean sea level. The river Alaknanda with its tributaries constitutes the drainage network in the area. Geologically, the region comprises the Lesser Himalayas and the Higher Himalayas. Structurally, the region is complex due to the presence of various thrusts, faults and intense deformations. The following datasets have been used to generate various thematic data layers: a) Remote Sensing data: IRS-1C LISS-III multi-spectral, PAN and stereo-PAN b) Survey of India toposheets at 1:50,000 scale (53 N / 2, 3, 6, 7) c) Geological Map d) Hill road design parameters (Indian Roads Congress 2001 recommendations) e) Field data on existing landslides, landuse/landcover and ground control points (using GPS surveys) The remote sensing data have been processed using ERDAS Imagine software. The BLUH software has been used for DEM generation from stereo-PAN image pair. The GIS analysis has been carried out using ILWIS software. A series of C++ programs have been written for route planning and the outputs have been suitably interfaced with the GIS derived themes. The IRS-1C LISS-III multi-spectral and PAN data have been co-registered to subpixel accuracy with the topographic map using a large number of ground control points. The LISS-III data have been corrected for atmospheric path radiance using the 'dark-object subtraction technique'. The Digital Elevation Model (DEM) of the study area has been generated using two methods. Firstly, an attempt has been made to generate a high resolution DEM from IRS- 1C stereo-PAN image pair using BLUH software (University of Hannover, Germany). However, the DEM generated from this method could not be used subsequently as it suffered from inaccuracies due to steep slopes, differential forest cover and snow. In view of the above, the Survey of India topographic map (1:50,000) has been digitised at 40 m contour interval, interpolated and resampled to generate the DEM. Slope, aspect and relative relief maps are then derived from the DEM using standard processes in raster GIS. Lithological and structural features have been extracted through digitisation of the geological map. The lineaments have been interpreted from 3x3 filter edge-enhanced LISS-III image and comparing the same with the geological map. At this stage, the structural features and lineaments are merged in a single layer and a distance buffer map is generated, which has been suitably reclassified. The drainage features have been digitised from topographic map and classified according to Strahler's ordering of streams. A drainage density map has also been generated. Landuse/landcover map has been generated by multi-source image classification of IRS-1C LISS-III image using logical channel approach. Multi-source classification has been adopted to reduce the effect of shadows cast by high mountain peaks in the region. Nine landuse/landcover classes that have impact on landslide activities in the region have been considered. These classes are: dense forest, sparse vegetation, agriculture, fallow land, barren land, settlements, fresh sediments, water body and snow. Normalised Difference Vegetation Index (NDVI) image and DEM have been included in the classification process to reduce the effect of shadows in the region and to improve the separability amongvarious classes. The most common classification algorithm - Maximum Likelihood Classifier - has been adopted to perform multi-source classification. It has been found that landuse/landcover map obtained from the combination of green, red and SWIR bands of LISS-III image, NDVI and DEM data layers produces the maximum overall accuracy of 92.06%. Post-classification filtering and editing has also been carried out to reduce the noise in the form of stray pixels in the map. The PAN-sharpened LISS-III image (based on IHS transform) together with substantial field observations has been used to produce a landslide distribution map. A new statistical approach for the preparation of LHZ map has been developed and applied for the region considering several thematic layers including topographic slope, aspect, relative relief, lithology, structure-buffers, drainage density and landuse/landcover. The weight factors have been calculated by overlaying the landslide distribution map on various thematic layers, and using a modified Landslide Nominal Hazard Factor (m-LNHF) method. The weighted thematic layers have been algebraically added to generate Landslide Hazard Index (LHI) map. Anew statistical procedure has been applied to classify the LHI map into various hazard zones. For calculating the cost of road development and maintenance, the thematic layers considered are: landslide occurrence map classified according to size, landslide hazard zones, higher order drainage (to consider bridge construction cost), landuse/landcover and lithology. The data layers have been integrated using an ordinal scale weighting-rating method. The thematic layers have been arranged in a hierarchical way, in order of increasing cost and aweight number (from 0to 9) is given to each layer. Similarly, each class within a layer has been given an ordinal rating. These weighted layers have been aggregated to generate a thematic cost map in which the pixel value implies the thematic cost to move through the pixel. The above thematic cost layer is a kind of isotropic cost layer and does not incorporate the costs due to neighbourhood distance and the terrain gradient, which are direction dependent. To consider these aspects, the neighbourhood schemes have to be used. In raster GIS, these schemes can be related to the movements similar to those in the game of chess, namely Rook's, Bishop's and Knight's patterns. In view of the steep topographic slopes of the area, the above existing neighbourhood schemes may not suffice and so two more possible moves (not existing in the game of chess) have been conceived and have been named here as Knight31 and Knight32. In this process, atotal of 32 unique neighbourhoods are possible. The neighbourhood distance between the source and connected neighbour on the 3-D surface is calculated using the Euclidean distance formula. The gradient between the connected neighbours is calculated and classified into various classes considering the limiting, ruling and exceptional gradient recommended by the Indian Roads Congress (2001) for hill roads. The gradient classes have been converted into gradient costs. The neighbourhood movement cost (NM-cost) is computed from neighbourhood distance, gradient and thematic cost map, and gives the cost to move in to a connected neighbour from a source pixel. Dijkstra's algorithm has then been used to find out the least-cost route between origin and destination points. The route thus generated has been converted into ILWIS readable segment map for its subsequent linking and integration with other thematic data layers. A dedicated software, named as Landslide Safe Intelligent Route Finder (LaSIRF) has been developed (in C++) to compute NM-cost and to implement the Dijkstra's algorithm for finding out the least-cost route. Since, the current version of the software developed is computationally very intensive, its performance has been examined on a set of example areas, each of 1.5 km x 1.5 km size, extracted from the original study area with different types of terrain conditions and different landslide distribution patterns. This has validated the concept and the approach developed in this research. The results indicate that the algorithm wisely avoids the landslide zones as well as higher cost areas, falling on its way from the origin to the destination points. The approach developed in this study can be successfully implemented for route planning in hilly regions that are susceptible to landslides.en_US
dc.typeDoctoral Thesisen_US
Appears in Collections:DOCTORAL THESES (Earth Sci.)

Files in This Item:
File Description SizeFormat 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.