★ An update to ECMWF's machine-learned weather forecast model AIFS
Updates ECMWF's AIFS with bounding layers for physical consistency, better precipitation and new variables; operational at ECMWF since February 2025.
Global medium-range NWP
Data-driven global weather models for 1 to 15 days, usually trained on reanalysis (ERA5), deterministic or with a single best estimate.
Search and filter this topic Plain list on GitHub
★ Highlights
★ AIFS -- ECMWF's data-driven forecasting system
AIFS is ECMWF's operational GNN-transformer forecast system, trained on ERA5 and operational analyses, with skilled upper-air, surface and tropical cyclone forecasts run publicly.
★ FengWu-GHR: Learning the Kilometer-scale Medium-range Global Weather Forecasting
First data-driven global model at 0.09 degree resolution, initialized by transfer from a low-resolution model, outperforming IFS-HRES in 2022 hindcasts.
China Meteorological Administration (CMA)Nanjing University of Information Science & Technology (NUIST)Shanghai Artificial Intelligence Laboratory
★ FuXi: A cascade machine learning forecasting system for 15-day global weather forecast
FuXi is a cascaded ML system giving 15-day global forecasts at 0.25 degrees, the first ML model to match the ECMWF ensemble mean at 15 days.
★ FengWu: Pushing the Skillful Global Medium-range Weather Forecast beyond 10 Days Lead
FengWu is a multi-modal transformer forecasting system that extends skillful forecasts to 10.75 days, beating GraphCast on 80% of 880 predictands.
Institute of Atmospheric Physics, Chinese Academy of SciencesNanjing University of Information Science & Technology (NUIST)Shanghai Artificial Intelligence Laboratory
★ GraphCast: Learning skillful medium-range global weather forecasting
GraphCast, a graph neural network trained on ERA5, makes 10-day global forecasts at 0.25 degrees in under a minute, beating operational systems on 90% of targets.
★ Pangu-Weather: A 3D High-Resolution Model for Fast and Accurate Global Weather Forecast
3D Earth-Specific Transformer trained on ERA5 is the first AI model to beat IFS across variables and lead times from 1 hour to one week.
★ FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators
FourCastNet uses adaptive Fourier neural operators to forecast global 0.25 degree weather in under 2 seconds, matching IFS at short leads and beating it for precipitation.
California Institute of Technology (Caltech)Lawrence Berkeley National Laboratory / NERSCNVIDIA
★ Improving data-driven global weather prediction using deep convolutional neural networks on a cubed sphere
CNN on a cubed-sphere grid with multi-step loss gives stable global forecasts for weeks, beating persistence, climatology and coarse NWP at short-to-medium range.
All papers (63)
AICON: An operational global machine learning weather forecasting model
AICON is a 13 km graph-neural-network global model run operationally at DWD, competitive with operational ICON for near-surface variables.
★ An update to ECMWF's machine-learned weather forecast model AIFS
Updates ECMWF's AIFS with bounding layers for physical consistency, better precipitation and new variables; operational at ECMWF since February 2025.
★ AIFS -- ECMWF's data-driven forecasting system
AIFS is ECMWF's operational GNN-transformer forecast system, trained on ERA5 and operational analyses, with skilled upper-air, surface and tropical cyclone forecasts run publicly.
★ FengWu-GHR: Learning the Kilometer-scale Medium-range Global Weather Forecasting
First data-driven global model at 0.09 degree resolution, initialized by transfer from a low-resolution model, outperforming IFS-HRES in 2022 hindcasts.
China Meteorological Administration (CMA)Nanjing University of Information Science & Technology (NUIST)Shanghai Artificial Intelligence Laboratory
Scaling transformer neural networks for skillful and reliable medium-range weather forecasting
Stormer, a plain transformer with randomized forecast training, beats leading models beyond 7 days on WeatherBench 2 using far less data and compute, and scales well.
Machine learning for numerical weather and climate modelling: a review
Chronological review of machine learning in weather and climate modelling, from solvers and parameterization emulation to full ML forecast models, with ethics and future directions.
Advancing Parsimonious Deep Learning Weather Prediction using the HEALPix Mesh
DLWP-HPX uses a HEALPix mesh and GRU-augmented U-Net with fewer variables at 110 km, reaching skill about one day behind Pangu-Weather and GraphCast at one week and stable one-year rollouts.
The Compatibility between the Pangu Weather Forecasting Model and Meteorological Operational Data
Case studies show Pangu-Weather runs stably from several operational NWP analyses as initial conditions, and better initial conditions improve its forecast skill.
Institute of Atmospheric Physics, Chinese Academy of Sciences
Deep learning and a changing economy in weather and climate prediction
Nature Reviews review of deep learning's impact on weather and climate prediction; abstract absent, judged from title and venue.
ECMWFETH ZurichMax Planck Institute for MeteorologyUniversity of Oklahoma (AI2ES)
★ FuXi: A cascade machine learning forecasting system for 15-day global weather forecast
FuXi is a cascaded ML system giving 15-day global forecasts at 0.25 degrees, the first ML model to match the ECMWF ensemble mean at 15 days.
Scaling Spherical CNNs
Scales spherical CNNs to large problems, reaching state of the art on QM9 and competitive results on several weather forecasting tasks; weather is a secondary target.
Spherical Fourier Neural Operators: Learning Stable Dynamics on the Sphere
Spherical Fourier Neural Operators avoid planar FFT artifacts and give stable year-long autoregressive atmospheric rollouts (1,460 steps) on the sphere.
Innovative Machine Learning Applications in Weather Forecasting and Long-Term Climate Analysis
Review of ML for weather prediction and climate analysis; claims a 97% accuracy model but gives no method detail to verify it.
M-ENIAC: A machine learning recreation of the first successful numerical weather forecasts
Physics-informed neural networks recreate the 1950 ENIAC barotropic vorticity forecasts on the sphere, achieving more accurate solutions than the original numerical solver.
Inductive biases in deep learning models for weather prediction
Reviews and analyses inductive biases (data, objectives, losses, architectures, optimisation) of deep-learning weather prediction models to guide more efficient, probabilistic designs.
★ FengWu: Pushing the Skillful Global Medium-range Weather Forecast beyond 10 Days Lead
FengWu is a multi-modal transformer forecasting system that extends skillful forecasts to 10.75 days, beating GraphCast on 80% of 880 predictands.
Institute of Atmospheric Physics, Chinese Academy of SciencesNanjing University of Information Science & Technology (NUIST)Shanghai Artificial Intelligence Laboratory
Physical Attention-Gated Spatial-Temporal Predictive Network for Weather Forecasting
PastNet uses attention-gated convolutions for spatiotemporal forecasting on synthetic and real datasets, outperforming baselines on weather-style prediction.
Retracted: A Systematic Review of Machine Learning and Artificial Intelligence Methods to Tackle Climate Change Impacts
Retracted review of machine learning and deep learning methods for weather prediction and IoT-based forecasting, with no specific results reported.
Maximising Weather Forecasting Accuracy through the Utilisation of Graph Neural Networks and Dynamic GNNs
Compares graph neural networks with traditional ML for multivariate weather forecasting from heterogeneous sensors; the abstract reports no quantitative results.
Machine Learning Methods to Weather Forecasting to Predict Apparent Temperature: A Review
Reviews machine learning methods as alternatives to physical models for weather forecasting, with a focus on predicting apparent temperature.
SwinRDM: Integrate SwinRNN with Diffusion Model Towards High-Resolution and High-Quality Weather Forecasting
SwinRDM couples a SwinRNN forecaster with a diffusion super-resolution step to reach 0.25° forecasts that beat IFS up to 5 days on Z500, T850, T2M and precipitation.
A modified deep learning weather prediction using cubed sphere for global precipitation
Modified cubed-sphere deep learning weather model predicts global daily precipitation from surface temperature, matching GFS skill at a 1-day lag.
Deep Neural Networks for Weather Forecasting
Deep networks map reanalysis fields at one time to a future time, a basic data-driven forecast setup with little methodological detail.
★ GraphCast: Learning skillful medium-range global weather forecasting
GraphCast, a graph neural network trained on ERA5, makes 10-day global forecasts at 0.25 degrees in under a minute, beating operational systems on 90% of targets.
Spatial Mixture-of-Experts
Spatial Mixture-of-Experts layer routes experts by location and sets new state of the art for medium-range weather prediction and ensemble post-processing.
★ Pangu-Weather: A 3D High-Resolution Model for Fast and Accurate Global Weather Forecast
3D Earth-Specific Transformer trained on ERA5 is the first AI model to beat IFS across variables and lead times from 1 hour to one week.
Generative Modeling of High-resolution Global Precipitation Forecasts
A GAN-enhanced FourCastNet better captures fine-scale structure and extreme global precipitation percentiles, matching NWP skill at 1-2 day leads.
University of California, BerkeleyLawrence Berkeley National Laboratory / NERSC
Inference from Real-World Sparse Measurements
A ViT-style transformer handles irregular sensor data directly, cutting wind nowcasting RMSE from 9.24 to 7.98 and beating GNN baselines across several domains.
Python-based Prediction of Rapid Intensification from MIMIC-TC Ensemble (PRIME)
Machine and deep learning models predict intensity of newly formed blocking events at onset with error comparable to GEFS, using less computation.
Multi-modal spatio-temporal meteorological forecasting with deep neural network
A multi-modal spatio-temporal deep neural network for meteorological forecasting; the abstract is missing, so the topic is inferred from the title.
Deep Learning augmented Weather Prediction based on Cubic Sphere for global precipitation forecasting
Dataset of inputs and targets for deep-learning-augmented cubic-sphere weather prediction for global precipitation forecasting.
★ FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators
FourCastNet uses adaptive Fourier neural operators to forecast global 0.25 degree weather in under 2 seconds, matching IFS at short leads and beating it for precipitation.
California Institute of Technology (Caltech)Lawrence Berkeley National Laboratory / NERSCNVIDIA
Forecasting Global Weather with Graph Neural Networks
A graph neural network steps the global atmosphere forward six hours at a time, achieving skill comparable to GFS and ECMWF at 1-degree resolution and running live on operational GFS data.
Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data
Deformable CNNs with transfer learning forecast North Atlantic-European weather regimes 1 to 15 days ahead, outperforming CNNs beyond 5-6 days lead time.
The AI Forecaster: Machine Learning Takes on Weather Prediction
Eos news piece on a CNN-based approach that produces exceptionally fast global weather forecasts from past weather data.
A Literature Review on Improvement of Weather Prediction By Using Machine Learning
Literature review of machine learning methods for improving weather prediction; no abstract available, so scope is inferred from the title.
Using neural networks to predict weather patterns and trends - challenges and opportunities
Reviews neural network methods for weather and climate prediction, comparing their effectiveness against numerical weather prediction models and discussing challenges.
Weather forecasting based on data-driven and physics-informed reservoir computing models
Data-driven and physics-informed reservoir computing models are proposed for weather forecasting. The abstract is unavailable, so the scope is uncertain.
Data-driven global weather predictions at high resolutions
Data-driven global model forecasts atmospheric state and 0.25-degree precipitation without physics, with training scaled to 1024 GPUs for rapid iteration.
Prediction of Synoptic-Scale Sea Level Pressure Over the Indian Monsoon Region Using Deep Learning
ConvLSTM predicts daily Indian monsoon sea level pressure anomalies at 7-day lead, with better synoptic skill than a conventional NWP model.
A Survey of Weather Forecasting based on Machine Learning and Deep Learning Techniques
Survey of machine learning and deep learning methods for weather forecasting, covering applications, types and problems; no new model.
Physics-informed generative neural network: an application to troposphere temperature prediction
PGnet is a physics-informed two-stage generative network that uses a mask to refine ERA5 500 hPa temperature predictions and reduce error accumulation in time-series rollouts.
Physics-informed neural networks for the shallow-water equations on the sphere
Physics-informed neural networks solve shallow-water equations on the sphere using a sequence of time-window networks and a boundary-encoding layer, tested on Williamson cases.
Can deep learning beat numerical weather prediction?
Discusses whether deep learning could fully replace numerical weather models and data assimilation, reviewing ML concepts and their fit for weather data.
Data-driven geophysical forecasting: Simple, low-cost, and accurate baselines with kernel methods
Learned-kernel kernel-flow emulators forecast weekly sea-surface temperature and daily North American temperature, beating LSTMs and classical baselines in accuracy and cost.
Numerical Weather Forecasting using Convolutional-LSTM with Attention and Context Matcher Mechanisms
A ConvLSTM encoder-decoder with CNN, attention and context matcher outperforms ConvLSTM, TrajGRU and U-Net on ERA5 and WeatherBench forecasts.
Application of Artificial Intelligence Technology to Numerical Weather Prediction
Reviews how artificial intelligence is applied across numerical weather prediction, including data assimilation, error correction, and end-to-end neural forecasting models.
On the similarities between Vision Transformers and Residual Networks for Weather Prediction
Compares Vision Transformers with residual networks for data-driven weather prediction; no abstract is available to confirm the findings.
Deep Learning-Based Weather Prediction: A Survey
Survey of deep learning approaches to weather prediction across forecast horizons (no abstract available; classification based on title).
Optimal Analysis of Weather Forecast Supporting Convolutional Neural Networks
CNN-based weather forecast analysis, a low-visibility journal paper (no abstract; classification based on title).
Deterministic weather forecasting models based on intelligent predictors: A survey
Survey classifying deterministic deep learning weather forecasting models and public datasets, with directions for future research.
Physics-inspired adaptions to low-parameter neural network weather forecasts systems
Spherenet convolution and hemisphere-specific network design improve WeatherBench CNN forecasts at 1.4 degrees for 500 hPa height and 850 hPa temperature up to day 10.
Data-driven medium-range weather prediction with a Resnet pretrained on climate simulations: A new model for WeatherBench
A residual CNN pretrained on climate model output and fine-tuned on reanalysis gives WeatherBench 5-day forecasts that outperform prior submissions and rival a physical baseline.
Geometric deep learning for medium-range weather prediction
Geometric (graph-based) deep learning for medium-range global weather prediction; no abstract is available, so details come from the title only.
A Machine-Learning-Based Global Atmospheric Forecast Model
Builds a low-resolution machine learning global forecast model that beats persistence for about five days and better matches humidity and temperature than a physics model.
★ Improving data-driven global weather prediction using deep convolutional neural networks on a cubed sphere
CNN on a cubed-sphere grid with multi-step loss gives stable global forecasts for weeks, beating persistence, climatology and coarse NWP at short-to-medium range.
Can Machines Learn to Predict Weather? Using Deep Learning to Predict Gridded 500‐hPa Geopotential Height From Historical Weather Data
CNNs forecast 500-hPa geopotential height up to 14 days, beating persistence and barotropic vorticity model but not the operational NWP model.
Deep Learning in Data Science
Short review of deep learning models for climate and weather prediction, claiming accuracy gains over statistical methods without quantitative results.
Artificial Intelligence May Be Key to Better Weather Forecasts
Perspective piece arguing that machine learning could turn weather data into faster, more accurate and more detailed forecasts.
Weather Forecasting using Machine Learning Techniques
Generic study applying machine learning techniques to weather forecasting; the abstract is unavailable, so details are unclear.
★ Challenges and design choices for global weather and climate models based on machine learning
Discusses challenges and design choices for deep-learning global weather forecasting, illustrated with a toy global neural forecast model.
Opportunities for Machine Learning in Weather Forecasting
AMS conference paper on opportunities for machine learning in weather forecasting; no abstract available, so content is inferred from the title.
Applying Neural Network Models to Prediction and Data Analysis in Meteorology and Oceanography
Early review of neural network applications to meteorological and oceanographic prediction and data analysis, from the pre-deep-learning era.
Also relevant (75)
★ Operational tropical cyclone forecasting with AI
Operational AI ensemble WeatherNext Cyclones predicts tropical cyclone track, intensity and size to 15 days, gaining about 1 day of lead time over leading operational models.
Extremes & hazardsColorado State University (incl. CIRA)Google DeepMindGoogle ResearchMet OfficeNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
★ AERIS: Argonne Earth Systems Model for Reliable and Skillful Predictions
AERIS is a 1.3 to 80B parameter Swin diffusion transformer reaching 10.2 ExaFLOPS on Aurora, stable to 90-day forecasts and outperforming IFS ENS.
Software, scaling & compressionAllen Institute for AI (Ai2, ACE)Argonne National LaboratoryUniversity of California, Los Angeles (UCLA)University of Chicago
FourCastNet 3: A geometric approach to probabilistic machine-learning weather forecasting at scale
FourCastNet 3 is a spherical-CNN probabilistic ensemble model that rivals diffusion models, forecasts 8-60x faster, and stays calibrated and spectrally realistic to 60 days.
Skillful joint probabilistic weather forecasting from marginals
FGN generates ensembles through learned model perturbations trained on per-location CRPS, achieving state-of-the-art probabilistic skill and skillful ensemble tropical cyclone tracks.
Building Machine Learning Limited Area Models: Kilometer-Scale Weather Forecasting in Realistic Settings
Kilometre-scale graph-based limited-area models for Denmark and Switzerland, driven by flexible boundary forcing, match or beat the NWP baseline for key surface variables.
★ Skilful global seasonal predictions from a machine learning weather model trained on reanalysis data
ACE2 produces skilful winter seasonal forecasts from SST-persisted initialisation, reaching NAO correlation 0.47 over 1993–2016, comparable to a physics-based model.
Sub-seasonal to seasonalAllen Institute for AI (Ai2, ACE)Met Office
★ AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score
AIFS-CRPS trains an ML ensemble model on a CRPS-based loss, outperforming the IFS ensemble for most medium-range variables and competing at subseasonal lead times.
GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations
GraphDOP is an ECMWF graph model trained and initialised only from satellite and conventional observations, giving skilful forecasts up to five days ahead.
Do data-driven models beat numerical models in forecasting weather extremes? A comparison of IFS HRES, Pangu-Weather, and GraphCast
Comparison of Pangu-Weather, GraphCast and IFS HRES finds data-driven models mostly beat IFS on global RMSE for 1-10 day forecasts and compete on temperature and wind extremes.
Continuous Ensemble Weather Forecasting with Diffusion models
Continuous Ensemble Forecasting samples diffusion ensemble trajectories in parallel without autoregressive rollout, achieving competitive global forecasts at arbitrary temporal resolution.
★ Regional data-driven weather modeling with a global stretched-grid
Graph-based stretched-grid model refines a global AI model to 2.5 km over the Nordics, beating the MEPS ensemble mean for 2 m temperature but underestimating extremes.
Regional & km-scale modelsECMWFMET Norway (Norwegian Meteorological Institute)
MetMamba: Regional Weather Forecasting with Spatial-Temporal Mamba Model
MetMamba uses a Mamba state-space backbone for regional forecasting, outperforming attention and neural operator backbones and enabling coupled limited-area training with a global host.
★ FuXi Weather: A data-to-forecast machine learning system for global weather
FuXi Weather is an ML system that assimilates multi-satellite observations and produces 10-day global forecasts, outperforming ECMWF HRES in observation-sparse regions like central Africa.
Data assimilationChina Meteorological Administration (CMA)Fudan University (FuXi)Nanjing University of Information Science & Technology (NUIST)Tsinghua University (NowcastNet)University of California, Los Angeles (UCLA)
★ Huge Ensembles Part I: Design of Ensemble Weather Forecasts using Spherical Fourier Neural Operators
A 1.1-billion-parameter SFNO ensemble with bred vectors and perturbed parameters passes spectral tests and is calibrated, setting up ML huge ensembles of 1,000+ members.
Probabilistic & ensemble forecastingLawrence Berkeley National Laboratory / NERSCNVIDIAUniversity of California, Berkeley
Leveraging data-driven weather models for improving numerical weather prediction skill through large-scale spectral nudging
Hybrid NWP-AI system spectrally nudges GEM toward GraphCast large scales, improving prediction skill and tropical cyclone tracks while keeping physically consistent fine scales.
Hybrid physics-ML & parameterizationEnvironment and Climate Change Canada (ECCC)
★ A Foundation Model for the Earth System
Aurora is a large Earth-system foundation model trained on over a million hours of data that beats operational systems for air quality, waves, cyclone tracks and high-resolution weather.
Foundation modelsThe Alan Turing InstituteMicrosoft Research (AI for Science)Silurian AIUniversity of Cambridge (Aardvark)
★ FuXi-ENS: A machine learning model for medium-range ensemble weather forecasting
FuXi-ENS, a VAE-based ensemble trained with CRPS and KL loss at 0.25°, beats ECMWF ENS in CRPS on 98.1% of 360 variable and lead time combinations.
Probabilistic & ensemble forecastingChina Meteorological Administration (CMA)Fudan University (FuXi)Nanjing University of Information Science & Technology (NUIST)
★ Aardvark weather: end-to-end data-driven weather forecasting
End-to-end model ingests raw observations to produce global and station forecasts, beating operational NWP on several variables using 8% of input data.
Data assimilationThe Alan Turing InstituteECMWFGoogle DeepMindMicrosoft Research (AI for Science)University of Cambridge (Aardvark)
Uncertainty quantification for data-driven weather models
Compares uncertainty quantification methods for Pangu-Weather ensembles over Europe; ML post-hoc UQ gives skill up to 5 days beyond ECMWF ENS.
Explainability, uncertainty & physicsKarlsruhe Institute of Technology (KIT)
YingLong-weather: AI-Based Limited Area Models for Forecasting of Non-precipitation Surface Meteorological Variables
YingLong is a 3 km AI limited-area model that beats WRF-ARW on surface wind speed in two domains while running much faster, using Pangu-Weather boundaries.
Regional & km-scale modelsBaiduChina Meteorological Administration (CMA)
★ A Practical Probabilistic Benchmark for AI Weather Models
Lagged ensembles give a parameter-free probabilistic comparison showing GraphCast and Pangu tie on CRPS, and that multi-step losses hurt probabilistic skill.
Probabilistic & ensemble forecastingLawrence Berkeley National Laboratory / NERSCNVIDIAUniversity of California, Berkeley
DiffDA: a Diffusion Model for Weather-scale Data Assimilation
DiffDA adapts the pretrained GraphCast network as a diffusion backbone to assimilate sparse observations at 0.25 degree, losing at most 24 hours of lead time versus state-of-the-art analyses.
Probabilistic Weather Forecasting with Hierarchical Graph Neural Networks
Graph-EFM is a hierarchical graph latent-variable model that generates arbitrarily large spatially coherent ensembles in one forward pass per step, matching deterministic error levels.
★ GenCast: Diffusion-based ensemble forecasting for medium-range weather
GenCast, a diffusion ensemble trained on reanalysis, beats ECMWF ENS on 97.4% of 1320 targets, with better tropical cyclone and wind power prediction, in 8 minutes.
Probabilistic & ensemble forecastingGoogle DeepMindGoogle Research
★ Towards an end-to-end artificial intelligence driven global weather forecasting system
Adas is an AI data assimilation model that, combined with FengWu, forms an end-to-end AI forecast system with skill comparable to IFS.
Data assimilationFudan University (FuXi)Nanjing University of Information Science & Technology (NUIST)Shanghai Artificial Intelligence Laboratory
★ FengWu-4DVar: Coupling the Data-driven Weather Forecasting Model with 4D Variational Assimilation
Couples the FengWu forecast model with 4D-Var using autodifferentiation instead of adjoints, producing cyclic analyses from simulated observations.
Data assimilationShanghai Artificial Intelligence LaboratoryTsinghua University (NowcastNet)
Do AI models produce better weather forecasts than physics-based models? A quantitative evaluation case study of Storm Ciarán
Case study of Storm Ciarán finds FourCastNet, Pangu, and GraphCast capture synoptic structure but underestimate peak winds and miss fine warning-relevant features.
★ Neural General Circulation Models for Weather and Climate
NeuralGCM couples a differentiable dynamical solver with ML components, matching ML and ECMWF ensemble skill for 1-15 day forecasts and stable multi-decade climate runs.
Hybrid physics-ML & parameterizationECMWFGoogle DeepMindGoogle ResearchMassachusetts Institute of Technology (MIT)
Can Artificial Intelligence‐Based Weather Prediction Models Simulate the Butterfly Effect?
Finds an AI weather model fails to reproduce rapid early growth of small initial perturbations, implying falsely unlimited predictability unless perturbations are large.
Constraining chaos: Enforcing dynamical invariants in the training of reservoir computers
Reservoir computers trained to enforce dynamical invariants gain longer, more stable forecasts with limited data on Lorenz 96 and a quasi-geostrophic atmosphere model.
Explainability, uncertainty & physicsNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Graph-based Neural Weather Prediction for Limited Area Modeling
Adapts graph-based neural weather prediction to limited-area modelling with a multi-scale hierarchical extension, validated on a Nordic regional model.
AI Foundation Models for Weather and Climate: Applications, Design, and Implementation
Review and perspective arguing that transformer-based weather and climate foundation models are mature enough to build, outlining criteria for downscaling, wildfire and hurricane tasks.
Foundation modelsIBM ResearchKarlsruhe Institute of Technology (KIT)NASA
On some limitations of data-driven weather forecasting models
Analysis of Pangu-Weather shows its forecasts lack the fidelity and physical consistency of physics-based models, suggesting part of its deterministic skill advantage comes from these peculiarities.
Evaluation of forecasts by a global data-driven weather model with and without probabilistic post-processing at Norwegian stations
Evaluates Pangu-Weather against ECMWF and MEPS at 183 Norwegian stations; probabilistic post-processing helps global models most, though MEPS remains best.
Post-processing & calibrationMET Norway (Norwegian Meteorological Institute)
★ WeatherBench 2: A benchmark for the next generation of data-driven global weather models
WeatherBench 2 provides an open evaluation framework, training and ground-truth data, and baselines for data-driven global 1-14 day forecasts.
Benchmarks, datasets & evaluationECMWFGoogle DeepMindGoogle Research
Deep Learning Improves GFS Wintertime Precipitation Forecast Over Southeastern China
DABU-Net, a data-augmented deep learning correction of GFS wintertime precipitation over southeastern China, cuts RMSE by 19–25% at 24–72 h.
Post-processing & calibrationTsinghua University (NowcastNet)Nanjing University of Information Science & Technology (NUIST)
Towards Stability of Autoregressive Neural Operators
Architectural fixes to neural operators curb autoregressive error growth, giving lower long-horizon errors on Navier-Stokes, shallow water and a global weather forecasting system.
Verification against in-situ observations for Data-Driven Weather Prediction
Releases a quality-controlled NOAA MADIS in-situ observation dataset to verify data-driven weather prediction models against real-world operational observations rather than reanalysis.
W-MAE: Pre-trained weather model with masked autoencoder for multi-variable weather forecasting
W-MAE pretrains a masked autoencoder on ERA5 and fine-tunes it for forecasting, lifting FourCastNet's anomaly correlation by about 20 percent.
A Machine Learning Outlook: Post-processing of Global Medium-range Forecasts
Neural-network post-processing of global 1 to 7 day forecasts across 30 levels improves RMSE by up to 12% for 850 hPa temperature at day 7, and discusses metric limits.
Post-processing & calibrationGoogle ResearchNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
★ ClimaX: A foundation model for weather and climate
ClimaX is a transformer foundation model pretrained on CMIP6 that fine-tunes to many weather and climate tasks, outperforming data-driven baselines.
Foundation modelsMicrosoft Research (AI for Science)University of California, Los Angeles (UCLA)code
A Deep Learning Method for Real-time Bias Correction of Wind Field Forecasts in the Western North Pacific
A multi-task double-encoder TrajGRU corrects ECMWF 10-day wind forecasts over the western North Pacific in real time, cutting speed and direction biases by 8-11% and 9-14%.
SWAR: A Deep Multi-Model Ensemble Forecast Method with Spatial Grid and 2-D Time Structure Adaptability for Sea Level Pressure
Deep multi-model ensemble method combining NWP outputs with window attention and a rolling recurrent network for the two-dimensional initial-lead time structure, applied to sea level pressure.
DL-Corrector-Remapper: A grid-free bias-correction deep learning methodology for data-driven high-resolution global weather forecasting
DLCR corrects and remaps FourCastNet gridded forecasts onto sparse, non-uniform observation locations, bias-correcting them against ground truth without gridded targets.
Data-driven Global Subseasonal Forecast Model (GSFM v1.0) for intraseasonal oscillation components
GSFM uses SE-ResNet to predict intraseasonal Z500 and T850 components, substantially outperforming CFSv2 at 10–30 day leads.
Sub-seasonal to seasonalNanjing University of Information Science & Technology (NUIST)
Short-range forecasts of global precipitation using deep learning-augmented numerical weather prediction
UNet post-processing of CFSv2 cuts global precipitation bias from +5–7 mm/day to within ±1 mm/day over land at 1–3 day leads.
Post-processing & calibrationNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of California, Los Angeles (UCLA)
SwinVRNN: A Data-Driven Ensemble Forecasting Model via Learned Distribution Perturbation
SwinVRNN learns VAE-based distribution perturbations within a Swin transformer recurrent model, giving better ensemble accuracy and spread than other perturbation methods on WeatherBench.
Probabilistic & ensemble forecastingAlibaba (DAMO Academy)Fudan University (FuXi)
Global Extreme Heat Forecasting Using Neural Weather Models
Convolutional neural weather models trained with extreme-focused losses improve global heatwave forecasts 1 to 28 days ahead, beating persistence and ECMWF S2S after two weeks.
Extremes & hazardsCalifornia Institute of Technology (Caltech)Google ResearchUniversity of Oklahoma (AI2ES)
Quantitative Precipitation Forecast Experiment Based on Basic NWP Variables Using Deep Learning
Deep learning model forecasts quantitative precipitation from basic NWP variables; no abstract available, so the method and gains are unverified.
Post-processing & calibrationChina Meteorological Administration (CMA)
★ Correcting Coarse‐Grid Weather and Climate Models by Machine Learning From Global Storm‐Resolving Simulations
Machine-learned nudging tendencies correct a 200-km FV3GFS model toward 3-km X-SHiELD, gaining 18 hours of 3-7 day skill and 30% better precipitation.
Hybrid physics-ML & parameterizationAllen Institute for AI (Ai2, ACE)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Washington
Dual-Branched Spatio-Temporal Fusion Network for Multihorizon Tropical Cyclone Track Forecast
DBF-Net fuses TC state vectors with reanalysis pressure fields to forecast Northwest Pacific tropical cyclone tracks at multiple horizons, improving accuracy over single-modality methods.
Probabilistic Forecasting with Generative Networks via Scoring Rule Minimization
Generative networks trained with prequential scoring rules, avoiding adversarial training, improve probabilistic calibration on chaotic models and a global weather benchmark.
Integrating Recurrent Neural Networks with Data Assimilation for Scalable Data-Driven State Estimation
RNNs replace the forecast model, error covariance and adjoint in the DA cycle, with DA updating hidden states; the approach scales via localization and parallelization.
ATMConvGRU for Weather Forecasting
ATMConvGRU adds axial attention memory to ConvGRU for sequential weather prediction of temperature, humidity, wind and radar echo, outperforming state-of-the-art methods.
Correcting Weather and Climate Models by Machine Learning Nudged Historical Simulations
A random forest learns nudging tendencies and is coupled to FV3GFS, reducing 500 hPa height forecast error and annual-mean precipitation RMSE by about 20% in year-long runs.
Hybrid physics-ML & parameterizationAllen Institute for AI (Ai2, ACE)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of WashingtonPrinceton University
A Hybrid Deep Learning Model for Improved Wind and Wave Forecasts
Deep learning improves wind forecasts from reanalysis, which then force a wave model, giving about 10% gains in wind and wave height over the Mediterranean and over 35% in the Aegean.
Combining distribution-based neural networks to predict weather forecast probabilities
Neural networks predict full probability density functions for WeatherBench geopotential and temperature at 3 and 5 days, stacked into a combined forecast that beats some NWP models.
Towards physically consistent data-driven weather forecasting: Integrating data assimilation with equivariance-preserving deep spatial transformers
Equivariance-preserving U-STN networks improve data-driven Z500 forecasts by 45% over U-Nets, and coupling them with ensemble Kalman data assimilation gives stable cycles under high observation noise.
Data assimilationLawrence Berkeley National Laboratory / NERSC
Stable Emulation of an Entire Suite of Model Physics in a State-of-the-Art GCM using a Neural Network
A single shallow neural network emulates the full atmospheric physics suite in NCEP GFS, remaining stable for medium-range forecasts and a year-long AMIP run with realistic output and speedup.
A Survey on Mathematical Aspects of Machine Learning in GeoPhysics: The Cases of Weather Forecast, Wind Energy, Wave Energy, Oil and Gas Exploration
Surveys machine learning in geophysics across weather forecasting, wind and wave energy, and oil and gas exploration, with notes on dataflow acceleration.
Machine learning emulation of gravity wave drag in numerical weather forecasting
Neural emulators of the non-orographic gravity wave drag scheme run stably to seasonal timescales, improve medium-range forecasts and run ten times faster on GPUs than the CPU scheme.
Hybrid physics-ML & parameterizationECMWFUniversity of Oxford
RainBench: Towards Global Precipitation Forecasting from Satellite Imagery
RainBench is a multimodal global benchmark combining simulated satellite data, ERA5 and IMERG precipitation, with the PyRain library and medium-range baselines.
RainBench: Enabling Data-Driven Precipitation Forecasting on a Global Scale
RainBench provides a global ML-ready dataset of precipitation and atmospheric variables for data-driven precipitation forecasting (no abstract; based on title).
Hybrid Weather Prediction: A Blend of Machine Learning and Numerical Modeling
AGU 2020 presentation on blending machine learning with numerical weather models; a conference abstract with no detailed results.
Hurricane Forecasting: A Novel Multimodal Machine Learning Framework
Hurricast combines encoder-decoder features with gradient-boosted trees for 24-hour tropical cyclone track and intensity, matching operational skill in seconds and improving NHC consensus.
Extremes & hazardsMassachusetts Institute of Technology (MIT)
Application of machine learning to improving numerical weather prediction
Neural network predicts and corrects COSMO regional model 2m temperature errors, improving forecasts in about 50% of cases.
★ WeatherBench: A benchmark dataset for data-driven weather forecasting
WeatherBench introduces an ERA5-derived benchmark with standard metrics and baselines (linear regression, deep learning, physical models) for data-driven medium-range global weather forecasting.
Benchmarks, datasets & evaluationECMWFTechnical University of Munich (TUM)University of Washingtoncode
★ WeatherBench: A Benchmark Data Set for Data-Driven Weather Forecasting
WeatherBench provides an ERA5-derived benchmark with metrics and baselines for data-driven global medium-range forecasting, enabling direct model comparison.
Benchmarks, datasets & evaluationECMWFUniversity of WashingtonTechnical University of Munich (TUM)code
STConvS2S: Spatiotemporal Convolutional Sequence to Sequence Network for Weather Forecasting
STConvS2S uses only convolutional layers for sequence-to-sequence weather forecasting of South American temperature and rainfall, 23% better than an RNN baseline and five times faster to train.
Tropical Cyclone Track Forecasting using Fused Deep Learning from Aligned Reanalysis Data
A neural network fusing past storm trajectories with reanalysis wind and pressure fields forecasts 24 h tropical cyclone displacement in seconds, complementary to dynamical models.
Extremes & hazardsInriaNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Toward Data‐Driven Weather and Climate Forecasting: Approximating a Simple General Circulation Model With Deep Learning
Deep network emulates a simple general circulation model, forecasting its states and producing stable climate runs with matching statistics and no long-term drift.
Fused Deep Learning for Hurricane Track Forecast from Reanalysis Data
Fused deep learning model forecasts hurricane tracks from reanalysis data; abstract unavailable, so classification is based on the title only.
Precipitation Data Assimilation System Based on a Neural Network and Case-Based Reasoning System
Integrates neural network and case-based reasoning into a precipitation data assimilation system to improve weather forecasts; results reported as effective.
Machine Learning Approaches to Reducing Tropical Cyclone Prediction Errors
An AMS 2018 conference abstract on applying machine learning to reduce tropical cyclone prediction errors. No abstract is available, so the assessment is low confidence.
Complex hybrid models combining deterministic and machine learning components for numerical climate modeling and weather prediction
Discusses hybrid models combining deterministic physics and machine learning components for climate modeling and weather prediction; abstract unavailable.
Hybrid physics-ML & parameterizationNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)