https://surrogatemodels.com/ News, papers, datasets, and funding and deals in Physics AI for engineers: Weather & climate. en https://surrogatemodels.com/papers/pangu-weatherhttps://surrogatemodels.com/papers/pangu-weatherThu, 12 Nov 2026 09:00:00 GMTA 3D transformer with Earth-specific priors (3DEST) that takes 13 pressure levels and the surface as one volume. Four networks are trained for lead times of 1, 3, 6 and 24 hours and chained so that a forecast needs few steps (hierarchical temporal aggregation), which reduces accumulated error.Weather & climate https://surrogatemodels.com/papers/gencasthttps://surrogatemodels.com/papers/gencastThu, 12 Nov 2026 09:00:00 GMTA conditional diffusion model: each 12-hour step is sampled by denoising, conditioned on the current and previous weather states. The denoiser encodes the grid onto a six-times-refined icosahedral mesh, processes it with a graph transformer and decodes back; sampling again gives another member of a 15-day ensemble.Weather & climate https://surrogatemodels.com/papers/neuralgcmhttps://surrogatemodels.com/papers/neuralgcmThu, 12 Nov 2026 09:00:00 GMTA general circulation model in which a differentiable spectral solver computes the large-scale dynamics and neural networks acting on single vertical columns predict the effect of unresolved processes such as clouds, radiation and precipitation. Solver and networks are trained together, end-to-end, on ERA5 trajectories of up to 5 days.Weather & climate https://surrogatemodels.com/papers/clifford-layershttps://surrogatemodels.com/papers/clifford-layersThu, 12 Nov 2026 09:00:00 GMTScalar and vector fields at each point are packed into one multivector, and convolutions and Fourier layers act on these multivectors with Clifford algebra, so the coupling between, for example, pressure and velocity components is built into each layer.Aero & CFDElectromagneticsWeather & climate https://surrogatemodels.com/papers/function-space-certificationhttps://surrogatemodels.com/papers/function-space-certificationMon, 09 Nov 2026 09:00:00 GMTInterval bound propagation and CROWN are formulated in function space: upper and lower bound functions are propagated through each neural-operator layer with rigorous numerical integration, and extended to higher-order derivatives to bound PDE residuals and solution error.Weather & climateGeneral PDEs https://surrogatemodels.com/papers/thuerey-neural-operator-resolvent-pnashttps://surrogatemodels.com/papers/thuerey-neural-operator-resolvent-pnasSun, 08 Nov 2026 09:00:00 GMTA neural network trained to emulate a system's dynamics is differentiated to give its eigenmodes and forcing-response modes, without the governing equations.Weather & climate https://surrogatemodels.com/news/physicsnemo-26-08https://surrogatemodels.com/news/physicsnemo-26-08Sat, 01 Aug 2026 09:00:00 GMTThe 26.08 release of NVIDIA PhysicsNeMo, the open-source framework for physics AI, moves the GeoTransolver and FLARE models out of the experimental namespace and adds experimental AeroJEPA, xDeepONet and Strata (weather emulation) models. On the data side it adds a Zarr backend, implicit-domain and exact-boundary mesh generation, and GPU remeshing with field transfer. The previous release, 26.05, focused on multi-dataset mesh training, data curation and active learning.Aero & CFDWeather & climate https://surrogatemodels.com/news/physicsnemo-diffusion-module-benchmarkshttps://surrogatemodels.com/news/physicsnemo-diffusion-module-benchmarksThu, 11 Jun 2026 09:00:00 GMTNVIDIA engineers Charlelie Laurent and Peter Harrington publish a comparison of `physicsnemo.diffusion` with equivalent pure-PyTorch diffusion pipelines. User-facing code falls to 59% of the baseline for training, 59% for inference and 46% for guided inference with diffusion posterior sampling, each a mean over 10 independent implementations. Training an 81M-parameter diffusion U-Net on 16-channel data on four GB200 GPUs, the module with compilation, bf16 mixed precision and Apex layers reaches 1.4x the baseline's throughput at 512x512; beyond that size the baseline runs out of memory, the optimised module reaches 1024x1024, and only multi-diffusion goes further, at lower throughput. Guided inference on one GB200 gains about 20% to 30% on domains under 256x256.Weather & climate https://surrogatemodels.com/news/physicsnemo-2-1https://surrogatemodels.com/news/physicsnemo-2-1Wed, 27 May 2026 09:00:00 GMTNVIDIA publishes PhysicsNeMo v2.1.0 on GitHub. A new unified external-aerodynamics recipe trains DoMINO, Transolver, GeoTransolver or FLARE on DrivAerML, SHIFT-SUV, HiLiftAeroML or a user's own data, mixing datasets at run time through mesh datasets and non-dimensionalisation, with surface or volume data. The release adds the GLOBE model with AirfRANS and DrivAerML examples, concrete-dropout uncertainty estimates and an out-of-distribution guard (OODGuard) for GeoTransolver, a `physicsnemo.sym` module that replaces the separate PhysicsNeMo-Sym package for PDE residuals, and a multi-diffusion subpackage for patch-based training and sampling. PyTorch 2.5 becomes the minimum version.Aero & CFDWeather & climate https://surrogatemodels.com/news/ecmwf-aifs-v2https://surrogatemodels.com/news/ecmwf-aifs-v2Tue, 12 May 2026 09:00:00 GMTECMWF implemented AIFS Single v2 and AIFS ENS v2 with the 06 UTC run of 12 May 2026, together with IFS Cycle 50r1. Both add a wave stream of 11 variables, the fraction of snow cover and a 10 hPa pressure level at the unchanged resolution of about 31 km. AIFS Single keeps its architecture and is fine-tuned on operational and IFS 50r1 analyses for 2018–2024; AIFS ENS replaces the afCRPS loss with a multi-scale loss. ECMWF reports significant wave height forecasts about 10% better than the operational IFS wave forecasts, and lists degradations in Arctic 2 m temperature and in Antarctic summer waves. With the same cycle it stopped the real-time runs of Aurora, FourCastNet, GraphCast and Pangu-Weather.Weather & climate https://surrogatemodels.com/news/physicsnemo-2-0https://surrogatemodels.com/news/physicsnemo-2-0Tue, 10 Mar 2026 09:00:00 GMTNVIDIA releases PhysicsNeMo v2.0.0, a refactor of the open-source framework that the 26.03 release notes describe as a major architectural change toward a modular toolkit. Reusable layers move to physicsnemo.nn, the base Module class to physicsnemo.core, and physicsnemo.launch is removed; a compatibility layer (physicsnemo.compat) keeps old import paths working, and a migration guide lists the changes. New packages are physicsnemo.diffusion (noise schedules, samplers, losses, multi-diffusion), physicsnemo.mesh (GPU mesh processing for simplicial meshes of any dimension) and refactored GPU datapipes. The 26.03 notes also add training recipes for crash dynamics with Transolver and MeshGraphNet, for car aerodynamics volume fields with GeoTransolver, and for StormScope with a diffusion transformer.Aero & CFDStructures & crashWeather & climate https://surrogatemodels.com/news/nvidia-earth-2-open-modelshttps://surrogatemodels.com/news/nvidia-earth-2-open-modelsMon, 26 Jan 2026 09:00:00 GMTAt the American Meteorological Society's annual meeting, NVIDIA released two weather models on Hugging Face and in its Earth2Studio package. Earth-2 Medium Range, built on the Atlas architecture, is a latent diffusion transformer that predicts 75 ERA5 variables on a 0.25-degree global grid in 6-hour steps, for forecasts up to 15 days; its model card lists a 2.5B-parameter main network and a 1.8B-parameter decoder. Earth-2 Nowcasting, built on StormScope, predicts GOES satellite and MRMS radar fields over the continental United States at 3 km and 10-minute steps, for zero to six hours ahead. A third model, HealDA, for global data assimilation, was announced for later in the year.Weather & climate