Uncertainty-Aware End-to-End AI Weather Forecasting: Disentangling Observation and Model Contributions
2026-08-31 • Machine Learning
Machine Learning
AI summaryⓘ
The authors improved a weather prediction model called Aardvark by adding ways to estimate uncertainty in its forecasts. They introduced two types of randomness: one that accounts for unpredictable noise in the original weather observations, and another that captures uncertainty in the model’s understanding of weather patterns. This approach helps explain where forecast differences come from and makes the model better and more transparent. Their method also shows better accuracy compared to the original deterministic model though it is still behind the top operational systems.
end-to-end weather forecastingaleatoric uncertaintyepistemic uncertaintyMonte Carlo dropoutdata assimilationensemble forecastingspread-skill ratioERA5 reanalysisCRPSdigital twins
Authors
Rodrigo Almeida, Noelia Otero, Jost Arndt, Simon Baur, Wojciech Samek, Jackie Ma
Abstract
End-to-end weather forecasting systems produce skillful global gridded and station forecasts directly from raw Earth observations, replacing the numerical weather prediction pipeline, including data assimilation, at a fraction of its cost. These systems are deterministic and issue no uncertainty. Here we render the Aardvark Weather model probabilistic by attaching one stochastic mechanism to each component: learned, input-dependent noise at the observation encoder, capturing aleatoric uncertainty inherited from the observing system, and Monte Carlo dropout in the processor, capturing epistemic uncertainty in the learned dynamics. The resulting nested ensemble attributes forecast spread to the two sources through a law-of-total-variance decomposition, cross-checked by withholding observation streams. Probabilistic finetuning significantly improves the mean forecast, by 4.2% on average across variables and lead times. The ensemble is calibrated against ERA5 through the medium range (spread-skill ratio 0.98), keeps station RMSE within 2.4% of the deterministic model while beating it in CRPS at every lead time, and trails the operational ECMWF ensemble. The encoder branch behaves as observation-driven uncertainty. Component-attributed uncertainty makes end-to-end forecasts more transparent, a step toward observation-driven digital twins of the atmosphere.