Abstract 

Landslides are major natural geological hazards andeach year these are responsible for enormous loss of human lives and property in Himalayan region 

spreading over Pakistan, India, Nepal, and Bhutan. Recent studies have revealed that landslides occur due to complex interaction ofseveral geo-environmental 

parameters such as lithology, geological structures(faults, lineaments), geomorphology, slope gradient, slope aspect, soil texture, soil type, drainage, land 

use and anthropogenic factors. Attempts have been made to integrate such factors based on either statistical or heuristic approach to produce landslide hazard zonation maps showing relative susceptibility of a given area to landslide hazards. However, such methods have several limitations and therefore, an attempt is made to integrate layers by training thedata set using artificial neural network (ANN) to arrive at more reliable results. The methodology was developed 

in an area within GiriValley in the Sirmour district of Himachal Pradesh, India.

Causative parameters and landslide maps were derived from interpretation of satellite images, topographic maps, field survey and other maps. These parameters were taken into consideration while using the back-propagation of neural network method. The weights obtained from the trained network were consequently utilized for map integration and classification. The resulting landslide susceptibility zonation map delineates the area into five classes: Very 

High, High, Moderate, Low and Very Low. These classes were validated by correlating the results with actual landslide occurrences. The early results are very encouraging and attempts are being made to further improve the training and classification results.