Snowfall Prediction Using Artificial Recurrent Neural Network (RNN)

Abstract

Prediction of weather is an attempt done  by meteorologists to forecast the weather conditions of an  area at some time in the future that may be expected. The parameters of the climatic condition are based on the  humidity, wind, temperature, rainfall and size of data set.  Snowfall has a tremendous effect on the livelihood and the  socio-economic development of the native individuals. It  extends the permanency of glaciers and thereby providing 

adequate water for drinking, irrigation and hydro-power  generation. Furthermore, it enhances business enterprise  and provides diversity advantages. Snowfall is a blessing  however if it is not correctly predicted, it can become a  curse. The cost of clearing the snow is high both in terms  of cost and time. Predicting snowfall for a short-term is  effective in scheduling when and how to clear the snow  and also necessary to decrease the cost of clearing it. Since  the weather is a non-linear phenomenon and its  randomness is very difficult to comprehend both in terms  of complexity and technology, deep-learning has been  applied as it is proficient in understanding the non-linearity  and also comparatively robust to perturbations. This paper  proposes a deep learning algorithm (LONG SHORT TERM MEMORY) which is employed on three different  datasets obtained from three different locations of Jammu  and Kashmir. The results found from this experiment  depicted that LSTM was capable of forecasting the snowfall with substantial accuracy.