Weather Forecasting using Deep Learning Algorithm

Authors

  • Keerthana P. J., Rajeswari P., Amirtharaj R., Pandiya Rajan G.

Abstract

Weather forecasting is significant for the early detection of weather impacts on various aspects of human life. Weather forecasting, for instance, supports autonomous vehicles in making decisions to reduce traffic accidents and gridlock, which are highly dependent on the sensing and prediction of external environmental factors such as rainfall, air visibility, and so on. The primary objective of meteorological scientists has always been to make accurate predictions correctly and promptly. Conventional theory-driven numerical weather prediction (NWP) methods, on the contrary, face significant challenges, including a lack of understanding of physical mechanisms, complexity in extracting useful knowledge, and the need for powerful computing resources. The successful application of data-driven deep learning techniques in various fields, including computer vision, speech recognition, and time series prediction, has demonstrated that deep learning techniques can effectively extract temporal and spatial features from spatiotemporal data. Weather data is an example of vast geospatial data. Deep learning-based weather prediction (DLWP) is expected to be a beneficial contributor to conventional practice. In this project, we implement deep learning-based weather forecasting, along with LSTM architecture design, spatial and temporal scales, as well as datasets, and benchmarks.

 

Downloads

Published

2022-04-28