Weather forecasting is an alarming challenge in the field of geo-sciences as it depends on several parameters which are dynamic and chaotic. Continuous changes in precipitation distribution are complicated, being effected by meteorological and geographical factors. The major forms of precipitation are rain, snow and hails which depends on atmospheric parameters and climatic conditions, in which snow is a significant component of the earth's hydrological cycle and a crucial factor of global and regional energy balance. In many areas of the world, snowmelt is of great importance for water supply of agricultural irrigation and people's daily life. Snow water equivalent, snow extent and melt onset are important parameters for climate models and hydrological models which are widely used in climate forecasting, flood controlling and irrigation management. Till date, snowstorms were measured using traditional radar based data, which face major problems such as attenuation issues with strong echoes, as their signals are weak enough. Hence, satellite images are one of the proficient sources in the identification of snow. In the process of image acquisition from the satellite imagery it would often find barriers like noise, burrs and so on, obscure or even cover the original image of an area or can reduce the image quality which include lot of noise. Therefore, in the present research Haar wavelet transform is adopted to enhance the image or to eliminate striping noise. Differentiation between rain and snow depends on the square root balance sparsity norm threshold value obtained on compressing and de-noising the satellite image. The proposed model yields an average accuracy of 83.07 % in the identification of snow.