weather model developers

Abstract: Climate emulators often enforce physical constraints by correcting each state predicted by a neural network, but evaluating only the corrected forecast conflates predictor skill with the effect of the correction. We introduce an evaluation framework that separates these contributions by comparing corrected forecasts, pre-correction predictions along corrected trajectories, and autonomous rollouts with correction disabled. We further cross correction during training and inference to distinguish its immediate effect from how it shapes the learned predictor, and use matched nonphysical controls to test whether improvements are specific to the physical correction. In ACE2, correction substantially improves physical consistency. In a three-seed ACE2 pilot, models trained with correction produce less accurate uncorrected precipitation forecasts at short horizons than models trained without correction, but more accurate forecasts at 10 and 30 days. This contradicts our preregistered hypothesis that the deficit would grow with forecast horizon. In a separate nine-seed experiment, directly supervising the uncorrected output reverses its precipitation forecast deficit at 72 hours while maintaining the accuracy of corrected forecasts at 24 hours. In ConCerNet's heat-equation system, projection-trained models have higher uncorrected trajectory-based vector-field RMSE than models trained without projection, while restoring the physical projection recovers accuracy far more effectively than a matched nonphysical control. Together, these results show that hard corrections can reshape the predictors trained alongside them, creating dependencies that are not visible from corrected forecasts alone. Hard-constrained emulators should therefore be evaluated by separately measuring predictor accuracy, correction effects, and their interaction during training.

Abstract: Ensemble weather forecasts can look reliable in global averages while hiding too little spread in some cases and too much in others. These errors vary across variables, spatial scales, lead times, and weather states. Adapting randomness during forecasting can correct these errors, but training fixes stochastic behavior beforehand and output calibration acts afterward. Inference guidance also acts during forecasting, but typically scores ensemble members separately rather than the ensemble as a whole. Reliability feedback arrives at the end, and tracing it backward through many chaotic forecast steps is expensive and unstable. We therefore introduce Dynamic and Informed Adjustment of the Native Stochastic Law (DIAL), which controls a frozen model's native stochasticity. DIAL learns from past forecasts to predict whether each variable and scale needs more or less spread. As a forecast develops, it estimates how possible noise changes will affect later forecast accuracy and reliability. It combines these predictions to make small noise updates at each step. This loop avoids retraining the weather model and full backpropagation through the forecast. We evaluate DIAL on FourCastNet 3 (FCN3) using later ERA5 years unseen during training. DIAL provides better scale-resolved reliability than learned postprocessing and reward-based control baselines. Relative to the unmodified Native FCN3 reference, DIAL achieves an 8.27% improvement in the primary multiscale continuous ranked probability score (CRPS).

Abstract: While AI weather models now rival operational forecasts, how they represent the atmosphere internally remains an open question: feature attribution reveals which input patterns matter, not what the model computes or how it combines information internally. We train sparse autoencoders (SAEs) on GraphCast to uncover its learned concepts, using atmospheric rivers as our phenomenon of focus. Both standard and Matryoshka SAEs show GraphCast computes atmospheric river intensity, measured by integrated vapor transport (IVT), as a stable internal variable, despite IVT being neither an input nor a target. In contrast to the unstructured concept retrieval of the standard SAE, the Matryoshka SAE orders concepts by importance and exposes their relations. Atmospheric river concepts persist across depth and direct interventions confirm causality. This method offers a way to find internal variables and determine which of them the model actually relies on, which is a prerequisite for asking whether those variables remain meaningful as the phenomenon changes under a warming climate.

Abstract: General circulation models (GCMs) are the primary tool for simulating planetary atmospheres. They play a vital role in understanding Mars's atmosphere, as forecasting its unique weather is mission-critical for operations such as entry, descent, and landing. Mars poses unusual challenges for these models, as observations are sparse compared to Earth. In addition, a thin \co{} atmosphere alongside a radiatively active dust cycle creates a volatile atmosphere with large diurnal temperature swings and no true terrestrial analog for validation. Existing Mars GCMs, including the LMD PCM, the NASA Ames Mars GCM, and PlanetWRF, are mature and physically detailed but are implemented in legacy Fortran with finite-difference or finite-volume solvers, and they do not expose gradients for calibration or machine learning. Here we present AEGIS, a modular differentiable Mars climate model that couples Mars's unique atmospheric physics to the Dinosaur dynamical core, with interfaces for neural closures. We showcase stable ten-Mars-year simulations that reproduce the seasonal \co{} cycle while conserving the total \co{} inventory, capture realistic large-scale surface-temperature structure, and produce surface pressure that follows Mars Orbiter Laser Altimeter (MOLA) topography. Gradients through coupled trajectories agree with finite differences and support physical calibration and neural training. We compare with conventional GCMs, highlighting the framework's computational efficiency and differentiability.

Abstract: Despite their high accuracy on point-wise metrics, machine learning weather forecasting models can exhibit different failure modes such as blurring, periodic irregularities, and other unphysical spatial artifacts. This has motivated a variety of metrics to detect known failure cases. Existing metrics fix a representation or transformation in advance, and that choice limits the artifacts they can detect. We propose to train a discriminator for separating reference data from the model's output, and using its output logit to obtain a divergence-like realism score. The discriminator learns whatever separates the model's fields from real weather, adapting to whichever failure mode that model exhibits. We compare our learned atmospheric critic to existing metrics using various synthetic corruptions applied to ERA5 reanalysis data. Our method successfully identifies the corruptions and ranks their severity, while existing metrics fail on at least one corruption. Additionally, we evaluate forecasts from real weather models, and find that the realism score degrades with longer lead times and the metric generally assigns higher realism to numerical models than to machine learning models.