https://ml4wx.github.io/Machine learning for weather & climate, tracked every weekhttps://ml4wx.github.io/papers/arxiv-2608.24651/arxiv-2608.24651Thu, 22 Oct 2026 08:00:00 +0000AICON is a 13 km graph-neural-network global model run operationally at DWD, competitive with operational ICON for near-surface variables.https://ml4wx.github.io/papers/doi-10.1038-s41586-026-10953-2/doi-10.1038-s41586-026-10953-2Thu, 22 Oct 2026 08:00:00 +0000Operational 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.https://ml4wx.github.io/papers/arxiv-2605.01126/arxiv-2605.01126Thu, 22 Oct 2026 08:00:00 +0000Extreme Weather Bench is an open community benchmark of case studies, observations and impact metrics for verifying AI and NWP models on high-impact weather.https://ml4wx.github.io/papers/arxiv-2604.16922/arxiv-2604.16922Thu, 22 Oct 2026 08:00:00 +0000ClimAgent is an LLM agent framework that runs end-to-end climate science analyses, introducing ClimaBench and a 40% gain over baseline LLM solutions.https://ml4wx.github.io/papers/arxiv-2511.20109/arxiv-2511.20109Thu, 22 Oct 2026 08:00:00 +0000A multi-agent LLM framework completes climate data workflows with 100% task completion on an 85-task benchmark, beating a GPT-5 baseline.https://ml4wx.github.io/papers/arxiv-2510.04017/arxiv-2510.04017Thu, 22 Oct 2026 08:00:00 +0000Zephyrus is an LLM agent with a weather-data environment and ZephyrusBench, outperforming text-only baselines by up to 44 points in correctness.https://ml4wx.github.io/papers/arxiv-2509.18994/arxiv-2509.18994Thu, 22 Oct 2026 08:00:00 +0000Updates ECMWF's AIFS with bounding layers for physical consistency, better precipitation and new variables; operational at ECMWF since February 2025.https://ml4wx.github.io/papers/arxiv-2509.13523/arxiv-2509.13523Thu, 22 Oct 2026 08:00:00 +0000AERIS 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.https://ml4wx.github.io/papers/arxiv-2509.12490/arxiv-2509.12490Thu, 22 Oct 2026 08:00:00 +0000SamudrACE is a coupled atmosphere-ocean ML emulator giving stable centuries-long runs at 1 degree with realistic ENSO variability absent in uncoupled mode.https://ml4wx.github.io/papers/arxiv-2507.12144/arxiv-2507.12144Thu, 22 Oct 2026 08:00:00 +0000FourCastNet 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.https://ml4wx.github.io/papers/arxiv-2507.05658/arxiv-2507.05658Thu, 22 Oct 2026 08:00:00 +0000HRRRCast is a ResNet-based emulator of the HRRR convection-allowing model over CONUS, with DDIM ensembles that beat HRRR at the 20 dBZ reflectivity threshold.https://ml4wx.github.io/papers/arxiv-2506.10772/arxiv-2506.10772Thu, 22 Oct 2026 08:00:00 +0000FGN generates ensembles through learned model perturbations trained on per-location CRPS, achieving state-of-the-art probabilistic skill and skillful ensemble tropical cyclone tracks.https://ml4wx.github.io/papers/arxiv-2505.06474/arxiv-2505.06474Thu, 22 Oct 2026 08:00:00 +0000cBottle is a generative diffusion emulator of global 5 km climate simulations and reanalysis on HEALPix, avoiding autoregressive rollout and reaching 12.5M-pixel resolution.https://ml4wx.github.io/papers/arxiv-2505.04802/arxiv-2505.04802Thu, 22 Oct 2026 08:00:00 +0000ORBIT-2 is a 10-billion-parameter ViT downscaling foundation model reaching 0.9 km global resolution on 65,536 GPUs with R² of 0.98–0.99 on 7 km benchmarks.https://ml4wx.github.io/papers/arxiv-2504.09340/arxiv-2504.09340Thu, 22 Oct 2026 08:00:00 +0000Kilometre-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.https://ml4wx.github.io/papers/arxiv-2503.23953/arxiv-2503.23953Thu, 22 Oct 2026 08:00:00 +0000ACE2 produces skilful winter seasonal forecasts from SST-persisted initialisation, reaching NAO correlation 0.47 over 1993–2016, comparable to a physics-based model.https://ml4wx.github.io/papers/doi-10.1038-s41467-025-56573-8/doi-10.1038-s41467-025-56573-8Thu, 22 Oct 2026 08:00:00 +0000Nature Communications review of AI for modeling and understanding extreme weather and climate events; abstract absent, judged from title and venue.https://ml4wx.github.io/papers/arxiv-2502.07532/arxiv-2502.07532Thu, 22 Oct 2026 08:00:00 +0000Diffusion-LAM is a conditional diffusion model that generates probabilistic limited-area forecasts from boundary conditions, demonstrated on the MEPS regional ensemble dataset.https://ml4wx.github.io/papers/arxiv-2412.15832/arxiv-2412.15832Thu, 22 Oct 2026 08:00:00 +0000AIFS-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.https://ml4wx.github.io/papers/arxiv-2412.15687/arxiv-2412.15687Thu, 22 Oct 2026 08:00:00 +0000GraphDOP is an ECMWF graph model trained and initialised only from satellite and conventional observations, giving skilful forecasts up to five days ahead.https://ml4wx.github.io/papers/arxiv-2412.12971/arxiv-2412.12971Thu, 22 Oct 2026 08:00:00 +0000ArchesWeather is a transformer deterministic model, and ArchesWeatherGen a flow-matching ensemble that beats IFS ENS and NeuralGCM on WeatherBench headline variables at academic compute cost.https://ml4wx.github.io/papers/arxiv-2412.11973/arxiv-2412.11973Thu, 22 Oct 2026 08:00:00 +0000A NeuralGCM-based hybrid model trained on satellite precipitation cuts climate-model precipitation biases and beats the ECMWF ensemble at mid-range forecasts.https://ml4wx.github.io/papers/arxiv-2412.04418/arxiv-2412.04418Thu, 22 Oct 2026 08:00:00 +0000Couples the ACE2 emulator to a slab ocean and trains it across CO2 levels, reproducing equilibrium warming but violating energy conservation in transient runs.https://ml4wx.github.io/papers/arxiv-2411.11268/arxiv-2411.11268Thu, 22 Oct 2026 08:00:00 +0000ACE2 is a 450M-parameter emulator that conserves dry air and moisture, runs about 1500 simulated years per day, and reproduces El Niño responses over 80 years.https://ml4wx.github.io/papers/arxiv-2411.05420/arxiv-2411.05420Thu, 22 Oct 2026 08:00:00 +0000WeatherGFM is a generalist weather foundation model using in-context visual prompting to handle up to ten weather tasks, including forecasting, super-resolution and post-processing.https://ml4wx.github.io/papers/doi-10.5194-gmd-17-7915-2024/doi-10.5194-gmd-17-7915-2024Thu, 22 Oct 2026 08:00:00 +0000Comparison 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.https://ml4wx.github.io/papers/arxiv-2410.19882/arxiv-2410.19882Thu, 22 Oct 2026 08:00:00 +0000Five recommendations for standardized, independent evaluation of ML-based Earth system models, stressing physical consistency for climate projections without historical analogues.https://ml4wx.github.io/papers/arxiv-2410.14932/arxiv-2410.14932Thu, 22 Oct 2026 08:00:00 +0000FourCastNet trained without Category 3-5 cyclones cannot forecast Category 5 storms, showing AI weather models fail to extrapolate to unseen extremes.https://ml4wx.github.io/papers/arxiv-2410.07472/arxiv-2410.07472Thu, 22 Oct 2026 08:00:00 +0000Systematic study of design choices for DL weather models finds fixed-grid architectures beat grid-invariant ones, proposes a hybrid, and shows multi-step fine-tuning is essential.https://ml4wx.github.io/papers/arxiv-2410.05431/arxiv-2410.05431Thu, 22 Oct 2026 08:00:00 +0000Continuous Ensemble Forecasting samples diffusion ensemble trajectories in parallel without autoregressive rollout, achieving competitive global forecasts at arbitrary temporal resolution.https://ml4wx.github.io/papers/arxiv-2410.01776/arxiv-2410.01776Thu, 22 Oct 2026 08:00:00 +0000Combining RCM dynamical downscaling with diffusion refinement lets large CMIP6 ensembles be downscaled cheaply, giving better uncertainty bounds than BCSD and smaller dynamical ensembles.https://ml4wx.github.io/papers/arxiv-2409.18529/arxiv-2409.18529Thu, 22 Oct 2026 08:00:00 +0000Shows weather-trained ML models forecast skillfully across pre-industrial, present-day and +2.9K climates, though two models show global-mean drift biases toward their training climate.https://ml4wx.github.io/papers/arxiv-2409.16247/arxiv-2409.16247Thu, 22 Oct 2026 08:00:00 +0000DLESyM, a deep learning Earth system model, simulates 1000-year observed climate without drift, matching CMIP6 variability at a fraction of the cost.https://ml4wx.github.io/papers/arxiv-2409.13598/arxiv-2409.13598Thu, 22 Oct 2026 08:00:00 +0000Prithvi WxC is a 2.3-billion-parameter MERRA-2 foundation model that transfers to forecasting, downscaling, gravity-wave parameterization and extreme-event tasks.https://ml4wx.github.io/papers/arxiv-2409.09371/arxiv-2409.09371Thu, 22 Oct 2026 08:00:00 +0000Introduces WeatherReal, a benchmark from global in-situ surface observations showing reanalysis-trained AI models diverge from station data for temperature, wind and precipitation.https://ml4wx.github.io/papers/arxiv-2409.02891/arxiv-2409.02891Thu, 22 Oct 2026 08:00:00 +0000Graph-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.https://ml4wx.github.io/papers/doi-10.1126-sciadv.adl3242/doi-10.1126-sciadv.adl3242Thu, 22 Oct 2026 08:00:00 +0000A CNN generates dynamically consistent counterfactual extreme heat events, estimating the 2023 southcentral North America heat was 1.18–1.42°C warmer due to warming.https://ml4wx.github.io/papers/doi-10.1038-s41558-024-02095-y/doi-10.1038-s41558-024-02095-yThu, 22 Oct 2026 08:00:00 +0000Nature Climate Change perspective on machine learning in climate modelling and analysis; abstract absent, judged from title and venue.https://ml4wx.github.io/papers/arxiv-2408.10958/arxiv-2408.10958Thu, 22 Oct 2026 08:00:00 +0000StormCast, a generative diffusion emulator of the 3 km HRRR convection-allowing model, gives competitive 1-6 hour radar reflectivity skill with realistic cold pools and power spectra.https://ml4wx.github.io/papers/arxiv-2408.09627/arxiv-2408.09627Thu, 22 Oct 2026 08:00:00 +0000Review of ML across climate physics, covering observation reconstruction, subgrid parameterization, causal inference and climate and weather prediction.https://ml4wx.github.io/papers/arxiv-2408.06400/arxiv-2408.06400Thu, 22 Oct 2026 08:00:00 +0000MetMamba 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.https://ml4wx.github.io/papers/arxiv-2408.05472/arxiv-2408.05472Thu, 22 Oct 2026 08:00:00 +0000FuXi 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.https://ml4wx.github.io/papers/arxiv-2408.03100/arxiv-2408.03100Thu, 22 Oct 2026 08:00:00 +0000A 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.https://ml4wx.github.io/papers/arxiv-2407.14158/arxiv-2407.14158Thu, 22 Oct 2026 08:00:00 +0000A diffusion model (CPMGEM) emulates 2.2 km UK convection-permitting precipitation from 60 km GCM inputs, reproducing extremes up to about 100-year return times at much lower cost.https://ml4wx.github.io/papers/arxiv-2407.06100/arxiv-2407.06100Thu, 22 Oct 2026 08:00:00 +0000Hybrid NWP-AI system spectrally nudges GEM toward GraphCast large scales, improving prediction skill and tropical cyclone tracks while keeping physically consistent fine scales.https://ml4wx.github.io/papers/arxiv-2407.00124/arxiv-2407.00124Thu, 22 Oct 2026 08:00:00 +0000A U-Net emulator of embedded cloud-resolving models gives a stable 5-year hybrid climate run with realistic cloud condensate and a 0.96 mm/day precipitation RMSE.https://ml4wx.github.io/papers/arxiv-2406.16947/arxiv-2406.16947Thu, 22 Oct 2026 08:00:00 +0000Score-based diffusion data assimilation at km scale assimilates 40 sparse weather stations, lowering held-out station RMSE by 10% versus HRRR.https://ml4wx.github.io/papers/arxiv-2406.11217/arxiv-2406.11217Thu, 22 Oct 2026 08:00:00 +0000WeatherQA is a multimodal benchmark of 8,000+ image-text pairs for severe convective weather reasoning; GPT-4o shows a substantial gap to human forecasters.https://ml4wx.github.io/papers/doi-10.1029-2023gl107377/doi-10.1029-2023gl107377Thu, 22 Oct 2026 08:00:00 +0000Shows Pangu-Weather, FourCastNet and GraphCast fail to reproduce sub-synoptic and mesoscale phenomena, lacking physical consistency versus physics-based models.https://ml4wx.github.io/papers/arxiv-2406.05019/arxiv-2406.05019Thu, 22 Oct 2026 08:00:00 +0000Backpropagation through GraphCast optimizes initial conditions, cutting 10-day Pacific Northwest heatwave errors over 90% and extending predictability to about 23 days.