Almost all weather forecasting systems combine analytical models and numerical methods to forecast future. In this work, I try to propose a new data driven and scalable method to model atmospheric processes to forecast weather.
I use a huge dataset of grib (GRIded Binary) files recorded from 1999 until now as input to model construction framework. They are downloaded from NCEP (National Centers for Environmental Prediction). Each file describes the present weather condition of the whole world as 2D grid. Every cell correspond to 1×1 (longitude*latitude) region and describe by about 240 to 350 parameters. Following files are extracted from this dataset and animate the evolution of temperature between 01/01/2000 to 20/01/2000 and 01/07/2000 to 20/07/2000.
1) Evolution of temperature between 01/01/2000 to 20/01/2000