ML-Based Weather Forecasting Models: A Comparative Study

Weather prediction plays a crucial role in our daily lives due to its impact in several sectors. In particular, it improves public safety by preparing for severe weather events and aids in various situations. Traditional weather forecasting, known as numerical weather prediction (NWP), relies on complex systems and mathematical models. Although these models provide high accuracy for weather prediction, they require expensive supercomputers and can take hours or even days to run. To cope with this, researchers are making advances in weather forecasting using machine learning (ML) to make robust predictions in a shorter time. In this paper, we first compare numerical weather prediction mod- els versus machine learning-based weather approaches. Then, we compare four major ML-based models: GraphCast [11], Pungu-Weather [2], Four- CastNet [12], and Fuxi [5] in terms of the most significant parameters.