57 papers in the archive.
Couples FourCastNet v2 boundary forcing with an 8 km ERF solver to simulate Hurricane Laura, reproducing rapid intensification to about 120 knots where AI-only models underestimate it.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
Provides a three-tier open protocol for evaluating physics-based, data-driven and hybrid climate models, withholding post-2015 observations and adding paleoclimate and perfect-model tests.
Benchmarks, datasets & evaluationAllen Institute for AI (Ai2, ACE)California Institute of Technology (Caltech)Columbia University (LEAP, Lamont-Doherty)Google ResearchLawrence Berkeley National Laboratory / NERSCMET Norway (Norwegian Meteorological Institute)Met OfficeNASANSF NCAR / UCARNVIDIANew York University (M2LInES)University of California, Los Angeles (UCLA)University of Chicago
Applies CGNet CNN to high-resolution CESM simulations, projecting AR frequency shifts and increased AR-linked extreme precipitation over western North America under RCP8.5.
Extremes & hazardsETH ZurichLawrence Berkeley National Laboratory / NERSCNSF NCAR / UCARNVIDIATomorrow.ioUniversity of Chicago
A Gaussian process surrogate calibrates SCREAM cloud parameters from a 153-member ensemble, cutting bias by more than 20% in 2-day simulations.
Hybrid physics-ML & parameterizationColumbia University (LEAP, Lamont-Doherty)Lawrence Berkeley National Laboratory / NERSCNASA
Uses 7424 SFNO hindcasts to show tropical cyclone tracks follow anomalous diffusion with ballistic regimes for some storms, informing track and landfall predictability over days to weeks.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCNVIDIAUniversity of California, Berkeley
Spherical Fourier Neural Operator produces 7,424 storyline forecasts of summer 2023, yielding heat extremes beyond NWP ensemble envelopes for 30% of land.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCNVIDIAUniversity of California, Berkeley
A Swin Transformer weather emulator shows predictable neural scaling under continual training, with IsoFLOP analysis identifying compute-optimal regimes for forecasting.
Global medium-range NWPLawrence Berkeley National Laboratory / NERSCOak Ridge National LaboratoryUniversity of California, Berkeley
Boosted trees with SHAP analyze 87 AmeriFlux sites and show radiation and soil moisture control carbon, water and energy fluxes, with VPD constraining water-use efficiency.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
Checkpoint-by-checkpoint analysis shows a Spherical Fourier Neural Operator learns tropical cyclone intensity for some storms, then unlearns it, linked to moist environments.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)Lawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
ML weather emulators trained on historical reanalysis reproduce ESM fast precipitation responses to CO2 changes without retraining, enabling study of fast radiative feedbacks.
Climate emulationAllen Institute for AI (Ai2, ACE)Lawrence Berkeley National Laboratory / NERSCNVIDIAStanford UniversityUniversity of California, Berkeley
SFNO-HENS and NeuralGCM hindcasts match ECMWF at week 3+ for North Pacific circulation and MJO, with realistic MJO propagation and teleconnections.
Sub-seasonal to seasonalColorado State University (incl. CIRA)Lawrence Berkeley National Laboratory / NERSCNVIDIAUniversity of California, Berkeley
Aerosol deep operator network emulates the E3SMv2 modal aerosol microphysics parameterization with high R-squared and fast CPU and GPU inference.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
Evaluates the NVIDIA SFNO huge ensemble on three 2023 North African heatwaves, forecasting 3 to 7 days ahead with good distribution tails.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCNVIDIA
Proposes a climate OSSE framework that uses machine learning to calibrate GISS-E3 physics against satellite data to value planetary boundary layer observations.
Observations & remote sensingCalifornia Institute of Technology (Caltech)Columbia University (LEAP, Lamont-Doherty)Lawrence Berkeley National Laboratory / NERSCNASANOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
SRCNN trained on dynamically downscaled precipitation emulates high-resolution downscaling, with quantile and exponential losses improving extremes and GCM applicability.
Downscaling & super-resolutionLawrence Berkeley National Laboratory / NERSCOak Ridge National LaboratoryUniversity of California, Berkeley
Argues that AI climate models should gain trust via component-level understanding, learning from dynamical model intercomparisons, not only post hoc XAI.
Explainability, uncertainty & physicsLawrence Berkeley National Laboratory / NERSCNSF NCAR / UCAR
Random forest and GAM on 54,000 soil carbon observations reveal controls that differ from CMIP6 Earth system models, guiding carbon-feedback representation.
Climate analysis & attributionIBM ResearchLawrence Berkeley National Laboratory / NERSCOak Ridge National LaboratoryTsinghua University (NowcastNet)
Five recommendations for standardized, independent evaluation of ML-based Earth system models, stressing physical consistency for climate projections without historical analogues.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)Lawrence Berkeley National Laboratory / NERSCNSF NCAR / UCAROak Ridge National LaboratoryUniversity of California, Berkeley
Nature Climate Change perspective on machine learning in climate modelling and analysis; abstract absent, judged from title and venue.
Climate emulationAllen Institute for AI (Ai2, ACE)California Institute of Technology (Caltech)Colorado State University (incl. CIRA)Columbia University (LEAP, Lamont-Doherty)Google ResearchInriaLawrence Berkeley National Laboratory / NERSCNSF NCAR / UCARNVIDIANew York University (M2LInES)University of California, BerkeleyUniversity of Oxford
Review of ML across climate physics, covering observation reconstruction, subgrid parameterization, causal inference and climate and weather prediction.
Hybrid physics-ML & parameterizationECMWFInriaLawrence Berkeley National Laboratory / NERSCUniversity of Chicago
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
A 7,424-member Spherical Fourier Neural Operator huge ensemble samples events 4 sigma from the mean and improves extreme-weather uncertainty quantification for summer 2023 forecasts.
Probabilistic & ensemble forecastingLawrence Berkeley National Laboratory / NERSCNVIDIAUniversity of California, Berkeley
Convolutional clustering autoencoders compress 70 years of GFDL-ESM2G CONUS projections by a factor of 660,000 and identify five physically reasonable climate zones, with clustering up to 9.2 times faster.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSCMassachusetts Institute of Technology (MIT)NASAUniversity of Chicago
Semi-supervised CNNs (ARCNNs) with style transfer detect atmospheric rivers across datasets, showing reanalysis latent heat transport varies considerably with detection uncertainty.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
A minimally modified SwinV2 trained on ERA5 beats IFS, and ablations on loss, model size and multi-step fine-tuning show which training choices matter.
Software, scaling & compressionLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
Machine-learning clustering of atmospheric river activity shows high-density clusters yield more high-category landfalls on the US West Coast, increasing with warming.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
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
ML emulators of the E3SM land model make variance-based sensitivity analysis of 19 wetland methane parameters feasible, cutting run time from 6 CPU hours to 0.72 ms.
Explainability, uncertainty & physicsLawrence Berkeley National Laboratory / NERSCMassachusetts Institute of Technology (MIT)
Hybrid model uses an ANN trained on analysis increments to correct errors in a coupled ocean-atmosphere model (MAOOAM), improving prediction skill across lead times.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
Review of ML for climate variability modes, extreme weather precursors, observation-model integration, downscaling and bias correction, aimed at domain scientists.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSCNSF NCAR / UCAROak Ridge National LaboratoryUniversity of California, Berkeley
Unsupervised local causal states detect hurricanes, atmospheric rivers and precipitation extremes in climate data, and track vortices in 2D turbulence.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCNVIDIA
Machine learning emulates the systematic error of a reduced-order coupled ocean-atmosphere model, and the resulting hybrid model beats the dynamical model alone in twin experiments, best when correcting both components.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
Interpretable attention-based fire model for African and South American burned area reveals lagged climate-wetness controls and projects South American burned-area recovery under high emissions.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSC
Hybrid network learns 4DVAR-style inversion directly from partial, noisy observations, using the variational mechanistic constraint for regularization and reduced computation.
Data assimilationLawrence Berkeley National Laboratory / NERSC
A GAN-enhanced FourCastNet better captures fine-scale structure and extreme global precipitation percentiles, matching NWP skill at 1-2 day leads.
Global medium-range NWPLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
FourCastNet uses adaptive Fourier neural operators to produce global medium-range forecasts about 80,000 times faster than NWP, scaling to thousands of GPUs.
Software, scaling & compressionLawrence Berkeley National Laboratory / NERSCNVIDIA
A convolutional VAE pretrained on imperfect quasi-geostrophic simulations and transfer-learned on noisy observations gives stable stochastic long-term rollouts with accurate climatology.
Climate emulationLawrence Berkeley National Laboratory / NERSCNVIDIA
ML and causal inference reduce CMIP6 land model evaporative fraction biases at 64 FLUXNET sites, raising correlation from 0.47 to 0.80.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSC
A DNN surrogate replaces the process-based wildfire model in E3SM, matching observed burned area with R2 0.93 and far lower parameterization cost.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
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.
Global medium-range NWPCalifornia Institute of Technology (Caltech)Lawrence Berkeley National Laboratory / NERSCNVIDIA
Couples reservoir-computing ML with an atmospheric GCM, improving forecasts for 7 to 8 days and giving a 10-year hybrid simulation with smaller systematic errors.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
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
Proposes machine learning to split seasonal and extreme precipitation into contributions from forcing, internal variability, and natural modes such as ENSO.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSCUniversity of California, Berkeley
Surveys physics-informed ML approaches with ten case studies for emulating, downscaling and forecasting weather and climate, reporting greater consistency and data efficiency.
Explainability, uncertainty & physicsCalifornia Institute of Technology (Caltech)Lawrence Berkeley National Laboratory / NERSCNVIDIAStanford UniversityUniversity of California, Berkeley
ClimateNet releases an expert-labeled dataset of tropical cyclones and atmospheric rivers in CAM5.1 output, and a deep learning segmentation model that enables pixel-level extreme event analysis.
Extremes & hazardsColorado State University (incl. CIRA)ETH ZurichLawrence Berkeley National Laboratory / NERSCNSF NCAR / UCARNVIDIAUniversity of California, Berkeley
Fusing rain radar images with NWP wind forecasts in a deep network improves 30-minute rain nowcasts, scoring 8% higher F1 than optical flow and 7% higher than radar-only.
Ensemble Kalman filter estimates model error from sparse noisy observations to train a neural parametrisation, yielding hybrid forecasts with better skill on Lorenz and a coupled ocean-atmosphere model.
Data assimilationLawrence Berkeley National Laboratory / NERSCUniversity of Reading
Spatial-transformer encoder-decoder autoregressively predicts two-layer quasi-geostrophic flow up to 9 days and stays stable for years, aiming at data-driven climate surrogates.
Climate emulationLawrence Berkeley National Laboratory / NERSC
Physics-based unsupervised representation learning segments extreme weather structures in CAM5.1 water vapor climate model output, a route to identifying events without labels.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSC
Position paper surveying how machine learning can help climate change mitigation and adaptation across sectors such as energy and disaster management; not focused on weather or climate science.
Applications & impactsETH ZurichGoogle DeepMindGoogle ResearchLawrence Berkeley National Laboratory / NERSCMassachusetts Institute of Technology (MIT)New York University (M2LInES)Stanford UniversityUniversity of California, Berkeley
Covariance-constrained GAN emulates Rayleigh-Bénard convection, an idealized atmosphere-like turbulent flow, as a closure model and cuts training time by up to 80%.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
Nature review on deep learning for data-driven Earth system science, covering weather, climate and process understanding; abstract absent, judged from title and venue.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSCMax Planck Institute for BiogeochemistryMax Planck Institute for Meteorology
Tests backward optimization and layerwise relevance propagation on a neural network for the Madden-Julian Oscillation, recovering known spatial structure and suggesting seasonality shifts.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)Lawrence Berkeley National Laboratory / NERSC
Physics-informed generative learning is proposed to emulate unresolved climate-model physics; the abstract is missing, so details are limited.
Hybrid physics-ML & parameterizationLawrence Berkeley National Laboratory / NERSC
Tiramisu and DeepLabv3+ segment extreme weather patterns and scale to 27,360 V100 GPUs at 325.8 PF/s sustained throughput.
Software, scaling & compressionLawrence Berkeley National Laboratory / NERSCNVIDIAOak Ridge National Laboratory
Abstract missing; applies topological data analysis with machine learning to classify atmospheric river patterns in large climate datasets.
Climate analysis & attributionLawrence Berkeley National Laboratory / NERSC
Introduces ExtremeWeather, a labeled climate dataset and semi-supervised spatiotemporal CNN for localizing hurricanes, extratropical cyclones and fronts in simulations.
Extremes & hazardsLawrence Berkeley National Laboratory / NERSCcode
