Artificial Neural Networks for Snow Avalanche Forecasting in Indian Himalaya

ASTRACT: Snow avalanches pose serious threat to Indian troops deployed in snow-bound areas of western Himalaya during winter months. The most viable way to mitigate avalanche threat in these areas is to precisely predict the time and place of their occurrences. Since, the factors involved in the formation of an avalanche are too many and underlying physical processes are quite complex, no prediction model can completely imitate the thought process and analysis methods of an expert forecaster. However, with the recent developments in the field of computational capabilities and Artificial Intelligence, improved models are coming up and proving to be reliable assistant to avalanche forecasters. In this direction, an attempt has been made to exploit the technique of Artificial Neural Networks (ANN). The base model is inspired by the popular non-parametric k-Nearest Neighbours method, where the two-parameter output is processed through ANN for classification. The model has been tested for an avalanche prone sector of Kashmir region and results are quite satisfactory. The paper discusses in detail the overall methodology adopted in the development of present model, and test results observed. Further possible use of Artificial Intelligence techniques in avalanche forecasting has also been discussed.