Abstract
Weather forecasting plays a vital role in today's society, from agriculture and logistics to predicting the output of renewable energies and preparing for extreme weather events. Deep learning weather forecasting models trained with the next state prediction objective on ERA5 have shown great success compared to numerical global circulation models. Here, we propose a methodology to leverage deterministic weather models in the design of probabilistic weather models, leading to improved performance and reduced computing costs. We design a probabilistic weather model based on flow matching, a modern variant of diffusion models, that is trained to project deterministic weather predictions to the distribution of ERA5 weather states. Our model ArchesWeatherGen surpasses IFS ENS and NeuralGCM on all WeatherBench headline variables (except for NeuralGCM's geopotential). Our work also aims to democratize the use of generative machine learning models in weather forecasting research.