Search papers, labs, and topics across Lattice.
This paper enhances the Aardvark Weather model by integrating stochastic mechanisms to account for both aleatoric and epistemic uncertainties, transforming it from a deterministic to a probabilistic forecasting system. By employing learned, input-dependent noise and Monte Carlo dropout, the model effectively disentangles the contributions of observations and model dynamics, leading to improved forecast accuracy. The probabilistic finetuning results in a 4.2% average improvement across various variables and lead times, while maintaining competitive performance against established operational ensembles.
Transforming a deterministic weather model into a probabilistic one reveals how uncertainty can significantly enhance forecasting accuracy and transparency.
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.
