SUMMARY
Application of neural networks is investigated for the prediction of avalanches on Chowkibal-Tangdhar road axis in Jammu and Kashmir. The networks are developed and trained monthwiseusing the past snow and weather parameters recorded at the Stage-II Observatory on the axisto generate an assessment of avalanche and non-avalanche activities. Two approaches havebeen considered for training the network. In the first approach, only avalanche activities observedin the axis were taken for training, and in the second approach, along with the observed activities,the opinion of expert forecasters were also considered. The performance of the networks variesfrom 67 to 82 per cent for correct predictions. Winter data for 2001-2002 has been used to validatethe network performance.