Home Articles Landslides in relation to terrain parameters – A Remote Sensing and GIS...

Landslides in relation to terrain parameters – A Remote Sensing and GIS approach


S. Sarkar
[email protected]


D.P. Kanungo
Geotechnical Engineering Division
Central Building Research Institute, Roorkee

Abstract
Landslide occurrence depends on the inter-play of several parameters and therefore it is imperative to know the contribution of these parameters on slope instability. In the present study, three terrain parameters drainage, lineament and road have been considered. The paper describes the utility of remote sensing and GIS for generation of these thematic layers and an integrated approach to evaluate layer-wise influence on landslide occurrence in parts of Darjeeling Himalaya.

Introduction
Landslide is the result of a wide variety of processes which include geological, geomorphological and meteorological factors. The important terrain factors are lithology, structure, drainage, slope, landuse, geomorphology and road network. A complete landslide hazard assessment requires an analysis of all these factors leading to instability in the region. The feature extraction of some of these factors can be done from the interpretation of satellite images. With the increase in efficient digital computing facilities, the digital remote sensing data and their analysis have gained enormous importance. Then the spatial and temporal thematic informations derived from remote sensing and ground based information need to be integrated for data analysis. This can be very well achieved using GIS which has the capabilities to handle voluminous spatial data. With the help of GIS, it is possible to integrate the spatial data of different layers to determine the influence of the parameters on landslide occurrence.

The present paper describes the generation of the four thematic layers; drainage, lineament, road and landslide. An attempt has been made to evaluate the relationship between these factors and landslides in the terrain using GIS technique.


Fig.1: Landslide map of Darjeeling Himalaya

Data
The data used to generate drainage, lineament, road and landslide maps of Darjeeling Himalaya is as follows :

  • IRS LISS III satellite data
  • IRS PAN satellite data
  • Topo sheets
  • Limited field data

Data Processing and Map Generation
All the four thematic layers using the above data are generated in GIS environment on 1:25,000 scale. The softwares used for this are ARC View GIS and ERDAS Imagine. The toposheets are initially registered with geographic lat-long coordinates. The PAN image is then co-registered with the geo-referenced toposheets by taking input ground control points from PAN image and reference ground control points (GCP) from registered toposheets. Then LISS-III image is co-registered with the geo-referenced PAN image. The registered images are then re-projected from geographic lat-long to polyconic projection so that the image co-ordinates are in meters. The merging of gray scale and multi-spectral images can enhance the satellite data interpretation because of both high spatial and spectral resolutions. Hence, a merged image from LISS III and PAN images is created through digital image processing. Feature extraction for all the four themes involves interpretation of individual LISS III, PAN and their merged product. The technique followed for map generation is described below


Fig.2: Drainage density map

Initially the PAN image is looked into and landslide slopes are very well interpreted. In this image the high contrast in the gray level due to thickly vegetated land and landslide affected barren slope is very well judged. However, the landslide areas with insignificant gray level contrast but having other spectral attributes are difficult to identify on PAN image. Hence in the next step, the merged PAN and LISS-III image is used. It is observed that landslide slopes are very nicely emerged as different identity even where it is not completely barren. All the identified landslides are mapped in the vector layer. The landslide vector layer is then exported to ARC View and landslide map is prepared after cleaning and building polygon topology. The theme table showing the area and perimeter of each landslide is automatically generated. In total 254 landslides are mapped. Few of these landslides are cross checked in the field also. The landslide map of the area is shown in the Fig. 1

Drainages from toposheets are digitised onscreen and then modified using satellite images. The vector file is then exported as shape file and the drainage map is generated in ARC View. Lineaments are the linear morpho-tectonic features of the terrain which include faults, fractures, ridges, major discontinuities etc. To extract these lineaments, satellite images are studied. It is observed that interpretation of merged data is more judicious than of individual image because of high spatial and spectral resolutions. The lineaments interpreted are digitised in vector mode and converted to shape file to generate the lineament map in ARC View. Considering road as one of the important anthropogenic factors inducing instability, a road map is generated showing only metalled roads of the study area. It is observed that the roads are very well identified on the merged image. All the roads are mapped initially from the toposheets and modified using the merged satellite data in vector form. After converting it to shape file, the road map is generated in ARC View.

Landslides in Relation to Terrain Parameters
The three terrain parameters drainage, lineament and roads are studied in relation to landslides. For this, drainage density, lineament density and proximity to roads are considered. A cell size of 250m is selected for computation of drainage and lineament density.

Landslide and drainage density
Drainage density is defined as the ratio of sum of the drainage lengths in the cell and the area of the corresponding cell. A drainage density map (Fig.2) is prepared after computing density for each cell using GIS. The values obtained range from 0 to 15.6 km/km2, which is finally classified into three classes of high, moderate and low drainage density. The spatial data of landslide and drainage density are then integrated in GIS to determine the frequency distribution of landslides for each density classes.


Fig. 3: Lineament density map

The landslide frequency which is actually the number of landslides per km2 of density class are given in Table 1.The trend obtained shows that landslide frequency is maximum in low drainage density class. This is not an usual trend in common practice. However, it could be possible that the areas of high drainage density reflecting high surfacial flow do not cause much instability as compared to areas of low drainage density where most of the water infiltrates causing instability. This statement can be supported by the fact that the terrain is dominantly of gneissic rocks and the weathered product of such rocks allows water to easily infiltrate.

Table 1. Landslide frequency in drainage density classes

Drainage density(km/km2) Area (km2) Number of landslides Landslide frequency
Low (0-5.2) 169.19 171 1.0107
Moderate (5.2-10.4) 86.75 77 0.8876>
High (10.4-15.6) 9.94 6 0.6036

Table 2. Landslide frequency in lineament density classes

Lineament density(km/km2) Area (km2) Number of landslides Landslide frequency
Low (0) 160.5 144 0.8972
Moderate (0-5) 96.0 96 1.0
High (5-13) 9.38 14 1.4925

Table 3. Landslide distribution in road buffers

Road buffer (m) Number of landslides
0-150 87
150-300 38
300-450 30
450-600 20
600-750 24
750-900 22

Landslide and lineament density
In a similar way, lineament density map is also prepared by computing density for each 250 m cell. The density values which range from 0 to 13 km/km2 is classified into three classes of low, moderate and high. The lineaments superimposed on density map are shown in the Fig.3. These density classes are then integrated with the spatial distribution of landslide to determine the landslide frequency in each lineament density class. The data are given in Table 2. This shows a direct relationship between landslide and lineament density which complements the fact that areas with more fracturing and faulting are always prone to landslide occurrence.

Landslide and proximity to road
As we all know that road construction most often causes slope instability, a distance of 150 m from the road on the uphill and downhill side is considered to study the influence of roads on landslide occurrence. For this, a road buffer map with 150 m interval is generated in GIS (Fig.4).


Fig. 4: Road Buffer Map

The number of landslides in each class of road buffer obtained by integrating the two layers is shown in the Table 3. It can be inferred from the table that maximum number of landslides is in the class of 150m and it decreases as the distance from road increases. However, it is observed that there is no significant change in landslide numbers after a distance of 450 m.

Conclusions
Landslide in relation to terrain parameters in Darjeeling Himalaya has been studied to decipher the influence of these parameters on landslide occurrence. It has been found that high spatial resolution satellite image (IRS PAN) and the merged product of PAN and LISS III are quite useful for terrain feature extraction. GIS is found to be a useful tool for preparation of thematic layers and spatial data analysis. The results obtained from the study has shown an inverse relationship between landslides and drainage density which may be due to high infiltration in weathered gneisses causing more instability in the area. Landslide frequency has been found to be maximum in high lineament density which indicates that the areas with more morpho-tectonic lineaments are more susceptible to landslides. The maximum landslide frequency within a 150m distance from the road strengthens the fact that the road is one of the important anthropogenic factors associated with landslide occurrence. Such type of study is useful for determining the relative importance of the terrain parameters and their categories for landslide hazard assessment.

Acknowledgement
Authors are grateful to the Director, CBRI for his kind permission to publish the work. The financial support from Ministry of Environment & Forests, New Delhi is acknowledged.