The paper introduces Climate Physics Dynamic Matching (ClimPhyDM), a variational, simulation‑free framework that blends an advection‑type physics prior with data‑driven components for weather forecasting. It leverages deep generative models to capture complex dynamical systems while preserving underlying physical structure. On the ERA5 benchmark, ClimPhyDM outperforms existing methods such as ClimODE and GB‑DM, achieving lower error over extended horizons and demonstrating improved temporal stability and resistance to error accumulation, all while training on a single modest 12 GB consumer GPU.
arXiv:2509. 15942v3 Announce Type: replace-cross Abstract: Internal variability is a dominant contributor to the uncertainty of predictions at the interannual to decadal timescale.
arXiv:2605.16929v2 Announce Type: replace Abstract: Global climate models are essential tools to simulate past and potential future pathways of climate change, as well as associated climate impacts....
arXiv:2601.21151v3 Announce Type: replace Abstract: Machine-learning approaches to weather forecasting often employ a monolithic architecture in which distinct physical mechanisms, such as advection,...
arXiv:2406. 14399v4 Announce Type: replace Abstract: The development of Time-Series Forecasting (TSF) models is often constrained by the lack of comprehensive datasets, especially in Global Station Weather Forecasting (GSWF), where existing datasets are small, temporally short, and spatially sparse.
arXiv:2608. 05103v1 Announce Type: new Abstract: Data assimilation (DA) uses Bayesian inference to update the state of a numerical forecast model with observed data.
arXiv:2602. 13416v2 Announce Type: replace Abstract: The proliferation of data-driven models in weather and climate sciences has marked a significant paradigm shift, with advanced models demonstrating exceptional skill in medium-range forecasting.
SimCast‑S2S is a generative latent‑diffusion model designed for probabilistic subseasonal‑to‑seasonal precipitation forecasting. It tackles three key challenges: it uses a diffusion pipeline to capture uncertainty, operates in a compact latent space to enable efficient large‑ensemble generation, and leverages transfer learning with low‑rank adaptation to train on limited reanalysis data after pretraining on climate simulations. The model outperforms deep‑learning baselines and competes with, or surpasses, operational systems such as the ECMWF‑S2S baseline without requiring extensive post‑processing.
Tianmu-TC is a physics‑constraints generative AI framework designed for global tropical cyclone forecasting. Trained on Western North Pacific data, it produces controllable outputs with reduced uncertainty, outperforming both deterministic and ensemble meteorological AI models as well as the ECMWF NWP system across global ocean basins. The model also demonstrates strong performance in challenging scenarios such as data sparsity, anomaly tracks, rapid intensification, and weakening, while maintaining significantly lower computational cost.
arXiv:2608. 05103v2 Announce Type: replace Abstract: Data assimilation (DA) uses Bayesian inference to update the state of a numerical forecast model with observed data.
SimCast‑S2S is a generative latent‑diffusion framework designed for probabilistic subseasonal‑to‑seasonal precipitation forecasting. It tackles three key challenges: it uses a diffusion‑based generative pipeline for uncertainty quantification, operates in a compact latent space learned by VAEs for efficient large‑ensemble generation, and employs transfer learning with LoRA to overcome limited training data. On reanalysis data, it outperforms deep‑learning baselines and competes with or surpasses state‑of‑the‑art operational systems such as ECMWF‑S2S.
This study introduces the first controlled benchmark of generative models for weather data assimilation using real station observations from 11,849 NOAA MADIS stations across the U.S. It evaluates key design choices—diffusion vs. flow matching, pixel vs. latent-space formulations, and inference-time conditioning strategies—against a classical 3D-Var baseline. The benchmark finds that learned generative priors and full-gradient guidance improve RMSE over ERA5, while other design variations offer minimal benefit, especially under sparse observation conditions.
arXiv:2607. 28220v1 Announce Type: cross Abstract: Atmospheric predictability declines rapidly beyond the next ten days, such that forecasts at longer lead times primarily convey large-scale trends rather than specific states.