An Envelope of Possibilities for Atlantic Basin Tropical Cyclone Seasons
Uses the ACE2 emulator ensembles to estimate the range of possible Atlantic hurricane counts for 2005, 2013 and 2024, finding 2024 underperformed in 90% of members.
Colorado State University (incl. CIRA)
199 papers in the archive.
Global Emissions Since the Paris Agreement Have Intensified Europe's Recent Heatwaves
A generative ML model shows post-2015 emissions raised Europe's summer temperature extremes with over 95% probability, lowering June 2025 heatwave temperatures by 0.34 °C.
Climate analysis & attributionColorado State University (incl. CIRA)Stanford University
Improving Global Precipitation Forecast using NASA Imerg in a Deep Learning Framework
Uses NASA IMERG satellite precipitation estimates in a deep learning framework to improve global precipitation forecasts. No abstract available.
Global medium-range NWPColorado State University (incl. CIRA)
★ 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)
Prithvi-Precip: Integrating Satellite Observations into an Atmospheric AI Foundation Model for Precipitation Forecasting
Prithvi-Precip fine-tunes the Prithvi-WxC foundation model for global precipitation, with satellite-derived training targets and direct satellite ingestion improving skill over MERRA-2 targets and GEOS forecasts.
Foundation modelsColorado State University (incl. CIRA)IBM ResearchNASAOak Ridge National Laboratory
U-Net-based spatial synergistic correction of gridded summer precipitation forecasts for CMA-MESO in Central China
A U-Net corrects gridded summer precipitation forecasts from the CMA-MESO model over Central China; the abstract is unavailable, so this summary relies on the title.
Post-processing & calibrationChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)NSF NCAR / UCARNanjing University of Information Science & Technology (NUIST)
Exploring the Potential of Machine Learning Post-Processing to Generate ERA5-Consistent Atmospheric Profiles from Geostationary Satellite Retrievals
LightGBM post-processing refines geostationary AMI humidity and temperature profiles toward ERA5, reducing mixing-ratio RMSE by 6–7% and RH by about 14%.
Observations & remote sensingColorado State University (incl. CIRA)
Identifying the Timing of Regional Summertime Minimum Temperature Threshold Crossings and the Potential Subsequent Climate Evolutions
CNNs with probabilistic outputs predict timing of summertime minimum temperature threshold crossings in Central North America; observation transfer learning shifts timing 4 to 10 years later.
Climate analysis & attributionColorado State University (incl. CIRA)
Machine Learning‐Based Swath Bias Correction for OCO‐3 Snapshot Area Mapping Mode Observations
Random forest identifies and corrects swath-based XCO2 biases in 19,839 OCO-3 SAM scenes, with 79% precision and 63% recall; about 9% of scenes flagged.
Observations & remote sensingCalifornia Institute of Technology (Caltech)Colorado State University (incl. CIRA)NASA
An Interpretable Deep Learning Model for Detecting Tropical Cyclone Genesis Through Incorporation of the Dynamic and Thermodynamic Environment
Interpretable 3D-residual GRU model detects tropical cloud clusters that will undergo genesis, reaching Heidke skill 0.80 and 94% detection; SHAP shows dynamics dominate.
Extremes & hazardsChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)Fudan University (FuXi)Nanjing University of Information Science & Technology (NUIST)
Global‐Frequency Synergy: A Novel Paradigm for Radar Echo Extrapolation via Attention and Fourier Convolution
GFST-LSTM adds attention and Fourier convolution to ST-LSTM for radar echo extrapolation, improving CSI at 40 dBZ by 22.9% over PredRNN.
NowcastingColorado State University (incl. CIRA)NSF NCAR / UCARNanjing University of Information Science & Technology (NUIST)
Precipitation diffusion downscaling and application to out-of-distribution simulations with and without stratospheric aerosol injection
A diffusion downscaler applied to CESM2 stratospheric aerosol injection runs suggests SAI could roughly halve the CONUS increase in annual maximum daily precipitation.
Downscaling & super-resolutionColorado State University (incl. CIRA)
The convergence of machine learning and data assimilation in Earth system science
Review arguing machine learning and data assimilation are converging, with ML accelerating DA and DA supplying uncertainty quantification and physical constraints.
Data assimilationBureau of Meteorology (Australia)Colorado State University (incl. CIRA)ECMWFMax Planck Institute for MeteorologyNASANOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)University of Reading
The Multi-University Consortium for Advanced Data Assimilation Research and Education (CADRE)
Describes the NOAA-funded CADRE consortium for data assimilation research and education, including novel machine-learning applications in DA.
Data assimilationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
Global Attention with Linear Complexity for Exascale Generative Data Assimilation in Earth System Prediction
STORM is a one-stage diffusion-based data assimilation framework with linear-complexity attention, scaling to 74,400 GPUs and 32,768-member ensembles.
Data assimilationColorado State University (incl. CIRA)Oak Ridge National Laboratory
Quantifying Very Extreme Precipitation and Temperature Using Huge Ensembles Generated by Machine Learning–Based Climate Model Emulators
A 10,560-year ACE2 ensemble shows that extreme value methods can estimate very rare precipitation and temperature quantiles over the contiguous US, including values beyond the training range.
Extremes & hazardsColorado State University (incl. CIRA)University of California, Berkeley
Reanalyses in the Age of Machine Learning: Why Dataset Curation Matters Now More than Ever
Essay arguing reanalysis should be reserved for assimilation-based datasets, since ML models trained on loosely defined reanalyses can inherit flaws.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Watch an AI Weather Model Learn (and Unlearn) Tropical Cyclones
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
Recent Weakening of the Global Radiative Feedback
A CNN trained on climate models and applied to observed temperature reconstructions shows the global radiative feedback weakened from about -3 to -2 W/m2/K since the mid-1990s, driven partly by Northeast Pacific warming.
Climate analysis & attributionColorado State University (incl. CIRA)Met Office
Distinguishing Between the Short‐Term Climate Responses to Different Stratospheric Aerosol Injection Latitudes With Explainable Artificial Intelligence
Neural networks with explainable AI identify stratospheric aerosol injection latitude from two-year temperature and precipitation responses, with temperature more distinguishable than precipitation.
Climate analysis & attributionColorado State University (incl. CIRA)
Deep-learning climate emulator ACE2 reveals a global decrease in tropical cyclone 5 frequency in the 15th Century under an El Niño-like sea surface temperature pattern
ACE2 emulator past-millennium runs link El Nino-like SST patterns to fewer global and Atlantic tropical cyclones via stronger wind shear and drier mid-troposphere.
Climate analysis & attributionColorado State University (incl. CIRA)Princeton University
Deep Learning for Multisatellite Precipitation Retrievals: Impact of Tomorrow.io’s Microwave Sounders
CNN fuses geostationary imagery and microwave sounder data for 10-minute, 4 km precipitation retrievals, gaining 30 to 50% skill over IMERG-Early against radar and gauges.
Observations & remote sensingColorado State University (incl. CIRA)Tomorrow.io
PDE foundation models are skillful AI weather emulators for the Martian atmosphere
PDE foundation model Poseidon, extended to 3D and fine-tuned on four Martian years, emulates Martian atmospheric weather with a 34.4% gain on a held-out year.
Foundation modelsColorado State University (incl. CIRA)IBM ResearchNASA
★ Subseasonal Forecasting and MJO Teleconnections in Machine Learning Weather Prediction Models
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
TCBench: A Benchmark for Tropical Cyclone Track and Intensity Forecasting at the Global Scale
TCBench is a standardized benchmark for 1-5 day global tropical cyclone track and intensity forecasts, comparing AI models (AIFS, GenCast, Pangu) against TIGGE; AI tracks are skillful, intensity less so.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)Columbia University (LEAP, Lamont-Doherty)ETH ZurichUniversity of California, BerkeleyUniversity of Oxford
Modulation of Tropical Cyclogenesis by the Convectively Coupled Kelvin Waves: Insights From Data‐Driven Climate Emulator ACE2
ACE2 emulator's 100-year ensembles show Kelvin waves robustly modulate tropical cyclogenesis, with genesis peaking one day after peak KW precipitation.
Climate analysis & attributionColorado State University (incl. CIRA)
★ Forced Component Estimation Statistical Method Intercomparison Project (ForceSMIP)
Intercomparison of statistical and machine learning methods for estimating forced climate response from single realizations; skilful in models but divergent when applied to observations.
Climate analysis & attributionColorado State University (incl. CIRA)ETH ZurichMax Planck Institute for MeteorologyNSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)New York University (M2LInES)Stanford UniversityUniversity of California, Los Angeles (UCLA)University of Oxford
Snow-to-Liquid Ratio Prediction over the Contiguous United States Using Machine Learning
Random forest snow-to-liquid ratio model applied to HRRR over CONUS cuts MAE to 2.92 versus 3.51–4.76 for operational methods.
Post-processing & calibrationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)Tomorrow.ioUniversity of Oklahoma (AI2ES)
Novel Multiview Machine Learning Classification of Snowflakes: Harnessing Convolutional Neural Networks and Multiangle Multicamera Instruments
Multiview CNN classifies snowflake shape, riming and melt state from multicamera images with 95.8–98.6% accuracy, supporting radar snow interpretation.
Observations & remote sensingColorado State University (incl. CIRA)
Enhancing Spaceborne Precipitation Radar Detail: Diffusion-Based Super-Resolution With Weak-Echo Recovery
Latent diffusion model super-resolves GPM spaceborne precipitation radar to 1-km scale, recovering weak echoes below sensitivity and sharpening convective structures in real cases.
Observations & remote sensingColorado State University (incl. CIRA)
An Improved Convolutional Neural Network for Bias Correction of Satellite Precipitation Products: Demonstration in Complex Terrain Regions Over Taiwan
Deep CNN bias correction of CMORPH and IMERG over Taiwan reduces mei-yu season relative bias from about -46% to 3% for CMORPH.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Enhanced Radar Quantitative Precipitation Estimation in South China Through Deep Learning Methods With Global Features
Graph neural networks outperform CNNs, MLPs and Z-R relations for radar quantitative precipitation estimation in South China, with the lowest MAE for extreme rainfall above 30 mm/h.
Observations & remote sensingChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)University of Oklahoma (AI2ES)
A Regional Benchmark for Deep Learning-Based Hourly Precipitation Nowcasting in Latin America
AINPP-PB-LATAM provides a Latin American satellite precipitation nowcasting benchmark (2018–2024) and baselines for five architectures, including AFNO, which excels at continuous skill.
Benchmarks, datasets & evaluationChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)NSF NCAR / UCAR
The Global Kilometric Tropical Cyclone Wind Database (CyclObs) Enabling High-Resolution Data-Driven Studies
CyclObs provides a global kilometre-scale tropical cyclone surface wind dataset from over 2,500 SAR observations across 500+ storms, intended for machine learning and model validation.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Winter Precipitation Type Diagnosis and Uncertainty Quantification with a Physically Consistent Machine Learning Method
An evidential neural network diagnoses winter precipitation type with calibrated uncertainty, beating area-based methods for freezing rain and ice pellets.
Post-processing & calibrationColorado State University (incl. CIRA)NSF NCAR / UCAR
Southern Hemisphere Surface Warming Drives Southwestern U.S. Precipitation according to AI-Informed Climate Model Simulations
Explainable AI on a CNN links south-central Pacific SST warming to Southwest US precipitation, with an AGCM experiment confirming the causal pathway via a shifted ITCZ and jet.
Climate analysis & attributionColorado State University (incl. CIRA)
CSU-PCAST: A Dual-Branch Transformer Framework for medium-range ensemble Precipitation Forecasting
CSU-PCAST, a Swin-transformer ensemble trained on ERA5 and IMERG, gives 30-member 15-day precipitation forecasts with lower CRPS than GEFS through 2023.
Probabilistic & ensemble forecastingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Multidata Causal Discovery for Statistical Hurricane Intensity Forecasting
Causal discovery selects predictors for statistical tropical cyclone intensity forecasts, improving SHIPS skill at 24–72 h lead times using ERA5-based data.
Application of an Interpretable Prototypical-Part Network to Subseasonal to Seasonal Climate Prediction over North America
ProtoLNet, an interpretable prototype network, forecasts North American subseasonal temperature from CESM2 and transfers prototypes to ERA5 with useful skill.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Forecasting the Future with Yesterday's Climate: Temperature Bias in AI Weather and Climate Models
FourCastNet, Pangu and ACE2 show cold land-temperature biases resembling climates 15-20 years earlier, with the largest errors in the hottest predicted temperatures.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)
Using machine learning to downscale coarse-resolution environmental variables for understanding the spatial frequency of convective storms
Pixel-based neural networks map coarse environmental variables to convection-permitting storm frequency with SSIM above 0.8, capturing diurnal and orographic patterns.
Downscaling & super-resolutionColorado State University (incl. CIRA)
Modulation of tropical cyclogenesis on subseasonal-to-interannual timescales in the deep-learning climate emulator ACE2
Evaluates whether the ACE2 deep-learning emulator produces physical tropical cyclogenesis, including realistic MJO modulation of genesis across basins on subseasonal to interannual scales.
Setting the Standard: Recommended Practices for Data Preprocessing in Data-Driven Climate Prediction
Position paper setting recommended practices for preprocessing inputs to AI/ML climate prediction models from subseasonal to decadal timescales, with case studies showing how choices change predictions.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
An Object-Based Evaluation of Output from a Deep Learning Model for Thunderstorm Nowcasting
Object-based tracking of ThunderCast thunderstorm nowcasts finds 69.4% of 17,054 tracks are true positives, and most of those lack lightning.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Projections of Lightning‐Ignited Wildfire Risk in the Western United States
CNN parameterizations of thermodynamic fields project more western US lightning days by mid-century, raising lightning-ignited wildfire risk on high fire-weather days.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
HRRRCast: a data-driven emulator for regional weather forecasting at convection allowing scales
HRRRCast 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.
Regional & km-scale modelsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
Knowledge-Guided Machine Learning: Illustrating the use of Explainable Boosting Machines to Identify Overshooting Tops in Satellite Imagery
Interpretable Explainable Boosting Machine detects overshooting tops in satellite imagery using human-guided texture features, trading some accuracy for transparency.
Observations & remote sensingColorado State University (incl. CIRA)
Uncertainty‐Aware Machine Learning Bias Correction and Filtering for OCO‐2: 1
Random forest bias correction for OCO-2 column CO2 retrievals reduces unphysical variability and agrees better with TCCON, with quantified uncertainty for operational use.
Observations & remote sensingCalifornia Institute of Technology (Caltech)Colorado State University (incl. CIRA)NASANSF NCAR / UCARUniversity of Oklahoma (AI2ES)
AI-informed model-analogs for understanding subseasonal-to-seasonal jet stream and North American temperature predictability
Neural-network-learned analog weighting improves S2S jet stream and US temperature forecasts over traditional analogs, climatology and persistence, including temperature extremes.
Sub-seasonal to seasonalColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Generative Algorithms for Wildfire Progression Reconstruction from Multi-Modal Satellite Active Fire Measurements and Terrain Height
A conditional GAN trained on WRF-SFIRE simulations estimates wildfire arrival-time progression from VIIRS and GOES data, achieving a 0.81 Sorensen-Dice overlap on five US fires.
Observation-based estimate of Earth’s effective radiative forcing
Machine learning links surface temperature to internal-variability radiation, giving an observation-based effective radiative forcing of 0.71 W m-2 per decade for 2001-2024.
Climate analysis & attributionColorado State University (incl. CIRA)
Measuring Sharpness of AI-Generated Meteorological Imagery
Proposes Gaussian Blur Equivalence to interpret sharpness metrics for AI meteorological imagery, demonstrated on the GREMLIN radar emulator and GraphCast outputs.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)NVIDIAUniversity of Oklahoma (AI2ES)
Tropical Cyclone Surface Winds From Aircraft With a Neural Network
SWANN neural network converts aircraft flight-level winds to tropical cyclone surface winds, improving on the operational wind-reduction procedure.
Observations & remote sensingColorado State University (incl. CIRA)NSF NCAR / UCAR
A Big Chill: Examining an Extremely Sharp Cold Front from Three Perspectives
A case study of an extreme 2022 High Plains cold front finds data-driven deep-learning forecasts better timed than physics models but missed the extreme temperature gradients.
Assessing MJO Tropical-Extratropical Teleconnections in Deep Learning Weather Prediction Models
Pangu-Weather and NeuralGCM reproduce MJO teleconnections two to three weeks out, and Pangu gives physically consistent PNA responses to MJO heating.
Climate analysis & attributionColorado State University (incl. CIRA)
Identifying Data Sources and Physical Strategies Used by Neural Networks to Predict TC Rapid Intensification
CNNs on satellite imagery add little skill over SHIPS scalars for 24-h tropical cyclone rapid intensification; XAI shows they rely on known features, with well-calibrated uncertainty.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
How to use score-based diffusion in earth system science: A satellite nowcasting example
Score-based diffusion models, with a CorrDiff variant best, improve geostationary satellite cloud nowcasts over U-Net and persistence and yield calibrated ensembles.
A Regimes‐Based Approach to Identifying Seasonal State‐Dependent Prediction Skill
Neural networks with self-organizing maps identify SST regimes that give skillful 7–8 week predictions of North Atlantic 500 mb height.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Sea Surface Salinity Provides Subseasonal Predictability for Forecasts of Opportunity of U.S. Summertime Precipitation
Neural networks trained on CESM2 Large Ensemble salinity anomalies give subseasonal skill for Midwest summer heavy rainfall at 0–56-day leads, verified with ERA5 moisture tracking.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Predicting Tropical Cyclone Track Forecast Errors using a Probabilistic Neural Network
A neural network predicts probabilistic tropical cyclone track uncertainty from NHC forecasts, better calibrated than static NHC error distributions and comparable to GEFS.
Post-processing & calibrationColorado State University (incl. CIRA)
★ A Comparison of AI Weather Prediction and Numerical Weather Prediction Models for 1–7-Day Precipitation Forecasts
Verification over CONUS finds GraphCast precipitation beats GFS and IFS on RMSE and equitable metrics, but NWP better captures precipitation intensity distributions and spatial placement.
Global medium-range NWPColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Advanced terrain-adaptive tropical cyclone wind field modeling using deep learning for infrastructure resilience planning
A deep learning model generates terrain-adaptive tropical cyclone wind fields for infrastructure resilience planning; abstract unavailable, so based on the title.
Convolutional Neural Networks Trained on Internal Variability Predict Forced Response of TOA Radiation by Learning the Pattern Effect
CNN trained on internal variability maps predicts forced TOA radiation response across climate models, with explainable AI showing the pattern effect at small scales.
Climate analysis & attributionColorado State University (incl. CIRA)
Reanalysis-based Global Radiative Response to Sea Surface Temperature Patterns: Evaluating the Ai2 Climate Emulator
Green's function experiments with the ACE emulator test its top-of-atmosphere radiative response to SST patterns, finding it likely underestimates historical warming feedbacks.
From bias to accuracy: Transforming satellite precipitation data in arid regions with machine learning and topographical insights
LSTM with elevation and temperature covariates corrects CMORPH and IMERG satellite rainfall over UAE mountains, improving NSE by about 45%.
Observations & remote sensingColorado State University (incl. CIRA)
Improving tropical cyclone rapid intensification forecasts with satellite measurements of sea surface salinity and calibrated machine learning
Calibrated machine learning with satellite pre-storm sea surface salinity improves tropical cyclone rapid intensification probability forecasts in several ocean basins.
Extremes & hazardsCalifornia Institute of Technology (Caltech)Colorado State University (incl. CIRA)NASA
Ensemble score filter with image inpainting for data assimilation in tracking surface quasi-geostrophic dynamics with partial observations
An ensemble score filter with image inpainting handles partial observations for data assimilation of surface quasi-geostrophic dynamics, a proof-of-concept on a toy atmospheric model.
Data assimilationColorado State University (incl. CIRA)Oak Ridge National Laboratory
Automated Tropical Cyclone Intensity Estimation Through Integration of Historical Information and Self-Diagnosis Using AI Models
Adding past intensities and a bias self-diagnosis step to satellite-based AI tropical cyclone intensity estimation cuts errors by 64.7% and a further 32.3%.
The Accuracy of Different Methods in Improving Quantitative Snowfall Estimation Based on Dual-Frequency Dual-Polarization Doppler Radar
Dual-polarization radar with CNN, LSTM and other ML models improves snowfall estimates; CNN reaches R2 of 0.98 for accumulation at one site.
Observations & remote sensingColorado State University (incl. CIRA)
Estimating Tropical Cyclone Intensity via Deep Learning and Storm Morphology Recognition From Satellite Imagery
Swin-Unet with storm-morphology loss terms estimates western North Pacific tropical cyclone intensity from Fengyun imagery, MAE 3.71 m/s.
PrecipSRGAN: A Machine Learning Model for Satellite Precipitation Downscaling
PrecipSRGAN uses a GAN with elevation data to downscale CMORPH satellite precipitation from 8 km to 4 km, reaching correlation 0.74 versus 0.41 for interpolation.
Downscaling & super-resolutionColorado State University (incl. CIRA)
Enhancing Weather Radar Reflectivity Emulation From Geostationary Satellite Data Using Dynamic Residual Convolutional Network
DRC-Net emulates surface radar reflectivity from geostationary cloud-top data, improving POD, FAR and CSI, including over ocean and mountains.
Observations & remote sensingColorado State University (incl. CIRA)
Combining climate models and observations to predict the time remaining until regional warming thresholds are reached
A CNN trained on climate model output, fine-tuned with observations via transfer learning, predicts when regional warming thresholds are reached, e.g. 2040 or earlier for 1.5 °C.
Climate analysis & attributionColorado State University (incl. CIRA)ETH ZurichStanford University
Can Ingredients-Based Forecasting Be Learned? Disentangling a Random Forest’s Severe Weather Predictions
Tree-interpreter analysis shows the operational CSU-MLP random forest severe-weather forecasts use ingredients consistent with severe-storm climatology.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Future seasonal surface temperature predictability with and without ARISE-stratospheric aerosol injection-1.5
Neural networks on ARISE-SAI-1.5 simulations show that stratospheric aerosol injection changes seasonal surface temperature predictability, especially in ENSO teleconnection regions in boreal winter.
Sub-seasonal to seasonalColorado State University (incl. CIRA)NSF NCAR / UCAR
Evaluation of Sub‐Hourly MRMS Quantitative Precipitation Estimates in Mountainous Terrain Using Machine Learning
Gradient-boosting error model maps sub-hourly MRMS radar precipitation errors across Colorado mountains, showing larger errors far from radars and at high elevation.
Observations & remote sensingColorado State University (incl. CIRA)
Predicting Cloud‐To‐Ground Lightning in the Western United States From the Large‐Scale Environment Using Explainable Neural Networks
CNNs with layer-wise relevance propagation predict western US cloud-to-ground lightning days (median AUC 0.8), usable for climate-model lightning parameterization.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)Princeton University
Errors of Opportunity: Using Neural Networks to Predict Errors in the Global Ensemble Forecast System (GEFS) on S2S Time Scales
Neural networks predict GEFS subseasonal forecast errors linked to MJO and BSISO phases, with explainability showing teleconnection-driven error sources.
Post-processing & calibrationColorado State University (incl. CIRA)
Recommendations for Comprehensive and Independent Evaluation of Machine Learning-Based Earth System Models
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
Gridded Severe Hail Nowcasting Using 3D U-Nets, Lightning Observations, and the Warn-on-Forecast System
A 3D U-Net nowcasts gridded severe hail up to 40 minutes ahead using Warn-on-Forecast and lightning data, beating HAILCAST baselines with max CSI near 0.48 at 20 minutes.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
Center-fixing of tropical cyclones using uncertainty-aware deep learning applied to high-temporal-resolution geostationary satellite imagery
GeoCenter uses uncertainty-aware deep learning on geostationary IR animations to locate tropical cyclone centers with 26.6 km mean error, matching microwave-based ARCHER-2 performance.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Evaluation of Tropical Cyclone Track and Intensity Forecasts from Artificial Intelligence Weather Prediction (AIWP) Models
Evaluates four open AI weather models on 2023 Northern Hemisphere tropical cyclones: track errors rival operational models, but intensity is biased low and worse than climatology.
Quality Control of Geostationary Lightning Mapper Observations for Tropical Cyclone Applications
Quality control methods, including ML-based anomaly detection, remove artifacts from GLM lightning near Atlantic tropical cyclones, lowering flagged lightning from 70% in 2017 to 2% in 2021.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Machine learning–based extreme event attribution
A 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.
Climate analysis & attributionColorado State University (incl. CIRA)Stanford University
Pushing the frontiers in climate modelling and analysis with machine learning
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
Distilling Machine Learning's Added Value: Pareto Fronts in Atmospheric Applications
Pareto-front model hierarchies compare semi-empirical and deep learning models for cloud cover, shortwave radiation and precipitation parameterizations, condensing ML gains into a ten-parameter equation.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)ECMWFNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)New York University (M2LInES)University of Oxford
The Chalmers Cloud Ice Climatology: retrieval implementation and validation
Chalmers Cloud Ice Climatology uses a convolutional quantile regression network to retrieve probabilistic global ice water path from geostationary IR back to 1980.
Observations & remote sensingColorado State University (incl. CIRA)
Validation of Cool-Season Snowfall Forecasts at a High-Elevation Site in Utah’s Little Cottonwood Canyon
Validates GFS and HRRR snowfall forecasts in Utah; a local random forest cuts snow-to-liquid ratio error from 4.5–7.7 to 3.7 but operational models underpredict precipitation.
Post-processing & calibrationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
The value of convergence research for developing trustworthy AI for weather, climate, and ocean hazards
Perspective from the AI2ES institute on convergence research for trustworthy AI in weather, climate and coastal hazards, integrating end-user trust assessment into AI development.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)NSF NCAR / UCARUniversity of Oklahoma (AI2ES)University of Washington
Incorrect computation of Madden-Julian oscillation prediction skill
Shows that a published ML MJO forecast skill was miscalculated; correct skill falls from 26–27 to 11–12 days, below dynamical models.
Sub-seasonal to seasonalColorado State University (incl. CIRA)University of Cambridge (Aardvark)
Improving Explainability of Deep Learning for Polarimetric Radar Rainfall Estimation
Deep network for polarimetric radar rainfall estimation paired with SHAP analysis showing how each radar variable drives rain-rate estimates across intensities.
Observations & remote sensingColorado State University (incl. CIRA)
Combining Neural Networks and CMIP6 Simulations to Learn Windows of Opportunity for Skillful Prediction of Multiyear Sea Surface Temperature Variability
Neural networks trained on large CMIP6 ensembles skillfully predict multiyear SST anomalies up to 5–7 years ahead and transfer to reconstructed observations.
Sub-seasonal to seasonalColorado State University (incl. CIRA)Stanford University
Identifying and Categorizing Bias in AI/ML for Earth Sciences
The authors propose a categorization of bias in AI for Earth system prediction, illustrating each bias type with examples across weather and climate tasks.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NSF NCAR / UCARUniversity of Oklahoma (AI2ES)University of Washington
StarNet: A Deep Learning Model for Enhancing Polarimetric Radar Quantitative Precipitation Estimation
A 3-D convolutional StarNet fuses polarimetric radar volumes with gauge labels and physical height priors to improve hourly rainfall estimates, reaching MAE 1.55 mm and CC 0.92.
Observations & remote sensingColorado State University (incl. CIRA)NSF NCAR / UCAR
Warm-Season Microwave Integrated Retrieval System Precipitation Improvement Using Machine Learning Methods
U-Net models correct MiRS satellite precipitation over the continental US, raising spatial correlation from 0.75 to 0.89 and reducing bias from 11.95% to -6.33%.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Airborne Radar Quality Control with Machine Learning
Random forest automates quality control of airborne Doppler radar data, classifying about 96% of weather and 93% of nonweather gates on withheld data.
Observations & remote sensingColorado State University (incl. CIRA)
Optimizing Seasonal‐To‐Decadal Analog Forecasts With a Learned Spatially‐Weighted Mask
A learned spatially weighted mask for analog selection improves seasonal-to-decadal forecasts of North Atlantic SST and ENSO over uniform analog methods on the MPI Grand Ensemble.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
A Novel CNN-Based Radar Reflectivity Retrieval Network Using Geostationary Satellite Observations
Proposes FR-CNN with feature redistribution and parallel attention to retrieve radar reflectivity from geostationary satellite data, outperforming U-Net on POD, FAR and CSI.
Observations & remote sensingColorado State University (incl. CIRA)
Exploring the Use of Machine Learning to Improve Vertical Profiles of Temperature and Moisture
A residual U-Net postprocesses RAP forecasts with satellite data, improving dewpoint profiles and CAPE by up to 40% with well-calibrated uncertainty for severe-weather forecasting.
Post-processing & calibrationColorado State University (incl. CIRA)
Machine Learning Estimation of Maximum Vertical Velocity from Radar
U-Nets retrieve storm maximum vertical velocity from 3D radar reflectivity with probabilistic outputs, reaching R² above 0.65 on WoFS simulations.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
Evidential Deep Learning: Enhancing Predictive Uncertainty Estimation for Earth System Science Applications
Evidential deep learning gives one-model aleatoric and epistemic uncertainty for winter precipitation type and surface flux regression, matching ensemble accuracy; releases the MILES-GUESS package.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)code
Deep Learning–Based Summertime Turbulence Intensity Estimation Using Satellite Observations
U-Net models estimate aviation turbulence intensity at three altitude layers from GOES-16 ABI imagery; U-Net3+ matches pilot reports best.
Observations & remote sensingColorado State University (incl. CIRA)
Generative Algorithms for Fusion of Physics-Based Wildfire Spread Models with Satellite Data for Initializing Wildfire Forecasts
A conditional Wasserstein GAN trained on WRF-SFIRE simulations infers wildfire arrival times from satellite detections, achieving a 0.81 Sorensen coefficient for fire perimeters.
Recent advances in seasonal and multi-annual tropical cyclone forecasting
Reviews seasonal and multi-annual tropical cyclone forecasting, including emerging ML/AI methods, ENSO-neutral predictability and landfall skill gaps relative to user needs.
Sub-seasonal to seasonalBureau of Meteorology (Australia)China Meteorological Administration (CMA)Colorado State University (incl. CIRA)Japan Meteorological Agency (JMA)Met OfficeMétéo-France (incl. CNRM)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Estimating Full Longwave and Shortwave Radiative Transfer with Neural Networks of Varying Complexity
U-Net++ and U-Net3+ networks emulate full shortwave and longwave RRTM radiation in GFSv16, running 7510 and 90 times faster with reliable heating rates.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Assessing decadal variability of subseasonal forecasts of opportunity using explainable AI
Neural networks with explainable AI identify decadal variability in tropically driven subseasonal precipitation skill in CESM2 and confirm the relationships in ERA5 reanalysis.
Sub-seasonal to seasonalColorado State University (incl. CIRA)NASANSF NCAR / UCAR
Leveraging spatiotemporal information in meteorological image sequences: From feature engineering to neural networks
A review of spatiotemporal feature engineering and CNN methods for meteorological image sequences, covering convection detection from satellite and solar forecasting.
Observations & remote sensingColorado State University (incl. CIRA)
Physical Analysis of the Impact of Polarization Parameters on Deep-learning Networks for Nowcasting
Permutation tests and feature visualization show dual-polarization parameters Zdr and Kdp improve deep learning radar nowcasts beyond 30 minutes, especially for strong convection.
Enhancing Spatial Variability Representation of Radar Nowcasting with Generative Adversarial Networks
Attentional GAN with a spatial variability loss reduces the blurry effect in radar nowcasts and outperforms competing models on the test set and storm cases.
NowcastingChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)Institute of Atmospheric Physics, Chinese Academy of SciencesTsinghua University (NowcastNet)
Using Machine Learning With Partial Dependence Analysis to Investigate Coupling Between Soil Moisture and Near‐Surface Temperature
CNNs with partial dependence analysis reveal nonlinear soil moisture–temperature coupling across mid-latitude regions using reanalysis data.
Climate analysis & attributionColorado State University (incl. CIRA)Stanford University
Quantifying “climate distinguishability” after stratospheric aerosol injection using explainable artificial intelligence
Explainable neural networks trained on Earth system model output quantify climate distinguishability after stratospheric aerosol injection, showing slower change under SAI.
Climate analysis & attributionColorado State University (incl. CIRA)
A Review of Machine Learning for Convective Weather
Reviews classical and deep learning methods for forecasting convective hazards such as tornadoes, hail, wind and lightning, and calls for trustworthy, operationally tested ML forecasts.
Extremes & hazardsColorado State University (incl. CIRA)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
Estimating full longwave and shortwave radiative transfer with neural
Neural networks emulate the full shortwave and longwave RRTM radiation scheme for GFSv16, running 6579x and 96x faster than the original with near-zero bias.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Trustworthy Artificial Intelligence for Environmental Sciences: An Innovative Approach for Summer School
Describes an AI2ES summer school teaching trustworthy and explainable AI for severe storms, tropical cyclones and space weather to 677 participants.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)NSF NCAR / UCARNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)University of Washington
Creating and Evaluating Uncertainty Estimates with Neural Networks for Environmental-Science Applications
Introduces six neural network uncertainty methods and evaluation scores for environmental science, demonstrated on dewpoint profiles and Himawari-8 convection prediction.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Improving E3SM Land Model Photosynthesis Parameterization via Satellite SIF, Machine Learning, and Surrogate Modeling
Boosted trees and random forests map satellite SIF to GPP to calibrate E3SM land photosynthesis parameters, reducing simulated GPP uncertainty after surrogate-based benchmarking.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)Oak Ridge National Laboratory
Data-driven predictions of the time remaining until critical global warming thresholds are reached
ANNs trained on climate model output estimate observed timing of 1.5°C and 2°C warming, with a 2033–2035 central estimate for 1.5°C.
Climate analysis & attributionColorado State University (incl. CIRA)Stanford University
Spring Regional Sea Surface Temperatures as a Precursor of European Summer Heatwaves
Explainable neural network on 100 MPI Grand Ensemble members identifies spring tripolar North Atlantic SST patterns as precursors of European summer heatwaves.
Climate analysis & attributionColorado State University (incl. CIRA)Max Planck Institute for Meteorology
Building a Fuel Moisture Model for the Coupled Fire-Atmosphere Model WRF-SFIRE from Data: From Kalman Filters to Recurrent Neural Networks
Builds a recurrent neural network fuel moisture model for WRF-SFIRE, initialized as a numerical ODE solver and trained on RAWS sensor and RTMA weather data to improve fire-weather inputs.
Identifying the regional emergence of climate patterns in the ARISE-SAI-1.5 simulations
Logistic regression and neural networks show stratospheric aerosol injection signals in regional temperature emerge within a decade, while precipitation signals are harder to detect.
Climate analysis & attributionColorado State University (incl. CIRA)Princeton University
Using Neural Networks to Learn the Jet Stream Forced Response from Natural Variability
Trains a CNN on internal variability of a long dry-core control run to predict the nonlinear jet stream response to tropospheric forcing, extending fluctuation-dissipation ideas beyond linear response.
Climate analysis & attributionColorado State University (incl. CIRA)NVIDIA
Polarimetric Radar Quantitative Precipitation Estimation Using Deep Convolutional Neural Networks
Deep CNNs with dense blocks estimate hourly rainfall from polarimetric WSR-88D radar, reaching MAE 1.58 mm and beating traditional Z-R relations.
Observations & remote sensingColorado State University (incl. CIRA)
Pixel-CRN: A New Machine Learning Approach for Convective Storm Nowcasting
Pixel-CRN nowcasts convective storms 30 and 60 min ahead from 3-D radar and reanalysis, and works with only a small training dataset.
Sinh-arcsinh-normal distributions to add uncertainty to neural network regression tasks: Applications to tropical cyclone intensity forecasts
Sinh-arcsinh-normal output distributions give neural networks asymmetric, heteroscedastic uncertainty, demonstrated on tropical cyclone intensity forecast errors.
Probabilistic & ensemble forecastingColorado State University (incl. CIRA)
Using Machine Learning with Partial Dependence Analysis to Investigate Land-Atmosphere Coupling
CNNs predict daily maximum temperature from soil moisture and geopotential, and partial dependence reveals nonlinear soil moisture-temperature coupling across mid-latitude regions.
Climate analysis & attributionColorado State University (incl. CIRA)Stanford University
An Interpretable Model of Climate Change Using Correlative Learning
A neural network trained on CMIP6 global temperature and precipitation means predicts the year, and Alopex-derived patterns reveal how climate change signals evolve.
Climate analysis & attributionColorado State University (incl. CIRA)
Internal variability and forcing influence model–satellite differences in the rate of tropical tropospheric warming
A machine learning framework trained on climate models attributes tropical tropospheric warming: forcing adds 0.25 K/decade, internal variability offsets 0.07, and an aerosol bias adds 0.04.
Climate analysis & attributionColorado State University (incl. CIRA)NSF NCAR / UCARUniversity of California, Los Angeles (UCLA)
Tools for Extracting Spatio-Temporal Patterns in Meteorological Image Sequences: From Feature Engineering to Attention-Based Neural Networks
Survey of feature engineering, convolutional, LSTM and attention methods for spatio-temporal patterns in meteorological image sequences, illustrated with solar forecasting and convection detection.
Observations & remote sensingColorado State University (incl. CIRA)
Attention-Based Scattering Network for Satellite Imagery
Scattering-transform features with channel attention from multi-band satellite imagery estimate tropical cyclone intensity and predict lightning occurrence.
GPROF-NN: a neural-network-based implementation of the Goddard Profiling Algorithm
Probabilistic neural GPROF-NN retrieves GPM surface precipitation, with a 3D CNN variant, improving GMI mean absolute error from 0.079 to 0.059 mm/h.
Observations & remote sensingColorado State University (incl. CIRA)
Carefully choose the baseline: Lessons learned from applying XAI attribution methods for regression tasks in geoscience
Shows XAI attributions for a climate-signal network change substantially with the chosen baseline, using large-ensemble SSP3-7.0 simulations, so baseline choice must match the science question.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Incorporating Uncertainty Into a Regression Neural Network Enables Identification of Decadal State‐Dependent Predictability in CESM2
Uncertainty-aware regression network in CESM2 finds low-uncertainty initial states for decadal sea surface temperature predictability linked to Atlantic and Pacific variability.
Climate analysis & attributionColorado State University (incl. CIRA)
★ A new paradigm for medium-range severe weather forecasts: probabilistic random forest-based predictions
Random forests trained on GEFS reforecasts and severe weather reports give probabilistic US severe-weather forecasts for days 4-8, comparable to SPC outlooks at days 4-5.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
A Primer on Topological Data Analysis to Support Image Analysis Tasks in Environmental Science
Tutorial introducing persistent homology and persistence landscapes for classifying mesoscale cloud organization in satellite images, with interpretable results and example code.
Observations & remote sensingColorado State University (incl. CIRA)
Comparison of Climate Model Large Ensembles With Observations in the Arctic Using Simple Neural Networks
Trains a shallow neural network to identify which climate model produced Arctic temperature maps, then applies it to reanalysis, using attribution to reveal model-specific regional patterns.
Climate analysis & attributionColorado State University (incl. CIRA)
Detection of Forced Change Within Combined Climate Fields Using Explainable Neural Networks
Explainable neural networks using layer-wise relevance propagation combine multivariate climate fields to detect forced climate change earlier than single fields.
Climate analysis & attributionColorado State University (incl. CIRA)
Quantifying the Effect of Climate Change on Midlatitude Subseasonal Prediction Skill Provided by the Tropics
Neural networks quantify how warming alters midlatitude subseasonal skill from tropical precipitation in CESM2 large ensemble, linked to seasonal variability changes.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Interpretable Climate Change Modeling With Progressive Cascade Networks
Progressive cascade networks start from linear models and add complexity only as data supports, then map global temperature and precipitation to years to reveal climate change patterns.
Climate analysis & attributionColorado State University (incl. CIRA)
Predicting Slowdowns in Decadal Climate Warming Trends With Explainable Neural Networks
Neural network with explainability predicts decadal global warming slowdowns from ocean heat content in CESM2 large ensemble, linking them to Interdecadal Pacific Oscillation-like patterns and testing on observations.
Climate analysis & attributionColorado State University (incl. CIRA)
Using Simple, Explainable Neural Networks to Predict the Madden‐Julian Oscillation
Shallow neural networks predict the real-time MJO index out to about 18 days in October–March, outperforming linear models and explained via attribution techniques.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Non‐Gaussian Detection Using Machine Learning With Data Assimilation Applications
SVM detects shifts in the Lorenz-63 variable distribution to switch between Gaussian and mixed Gaussian-lognormal 3DVar, reducing analysis error.
Data assimilationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Machine learning–based observation-constrained projections reveal elevated global socioeconomic risks from wildfire
Uses a machine learning framework to observationally constrain 13 CMIP6 Earth system models, projecting higher socioeconomic wildfire exposure in Africa despite lower fire carbon emissions.
Extremes & hazardsColorado State University (incl. CIRA)Oak Ridge National Laboratory
Can we integrate spatial verification methods into neural-network loss functions for atmospheric science?
Spatially enhanced loss functions using neighbourhood and spectral filters improve neural network convection forecasts, with spectral-filter models giving better-calibrated probabilities than pixelwise training.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Investigating the fidelity of explainable artificial intelligence methods for applications of convolutional neural networks in geoscience
Benchmarks XAI attribution methods on a CNN that counts atmospheric rivers in climate simulations, exposing fidelity issues such as gradient shattering and sign ambiguity.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
An Improved Deep Learning Model for High-Impact Weather Nowcasting
A self-attention GRU nowcasts radar echoes for landfalling hurricanes and extreme convective rainfall in the southeast US, aiming to improve skill beyond the first hour.
NowcastingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Rainfall Estimation From TEMPEST-D CubeSat Observations: A Machine-Learning Approach
An artificial neural network maps CubeSat TEMPEST-D microwave brightness temperatures to surface rainfall, reaching 0.78 structural similarity on independent storms against MRMS radar.
Observations & remote sensingColorado State University (incl. CIRA)
Low Cloud Detection in Multilayer Scenes Using Satellite Imagery with Machine Learning Methods
Random forest and neural network models on GOES-ABI imagery improve low-cloud detection in multilayer scenes, raising probability of detection from 0.685 to 0.815 versus the CCL product.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
The Need for Ethical, Responsible, and Trustworthy Artificial Intelligence for Environmental Sciences
Position paper arguing AI in environmental sciences risks bias and injustice, with weather and climate examples; calls for responsible AI practice.
Applications & impactsColorado State University (incl. CIRA)NSF NCAR / UCARUniversity of Oklahoma (AI2ES)University of Washington
Integrated causal-predictive machine learning models for tropical cyclone epidemiology
Causal-predictive ML links tropical cyclone exposure to mortality and hospitalization in Medicare recipients, finding substantial respiratory hospitalization increases with wind speed as main driver.
Applications & impactsColorado State University (incl. CIRA)Columbia University (LEAP, Lamont-Doherty)
Improving Nowcasting of Convective Development by Incorporating Polarimetric Radar Variables Into a Deep‐Learning Model
Adding polarimetric radar variables KDP and ZDR to a deep-learning nowcasting model raises CSI by 13.2% at 30 min and 17.4% at 60 min.
NowcastingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Developing Deep Learning Models for Storm Nowcasting
Systematic study of deep learning storm nowcasting design choices (history length, loss, predicted frames) from radar data, yielding models that outperform existing implementations.
Adding Uncertainty to Neural Network Regression Tasks in the Geosciences
Neural network regression outputs a general probability distribution for heteroscedastic, asymmetric uncertainty, demonstrated on synthetic climate data.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Deep Learning for Bias Correction of Satellite Retrievals of Orographic Precipitation
CNN bias correction of CMORPH satellite precipitation over US coastal mountains raises correlation from 0.55 to 0.88 against Stage IV gauge-radar data.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Deep Learning for Polarimetric Radar Quantitative Precipitation Estimation during Landfalling Typhoons in South China
QPENet deep learning on dual-polarization radar, trained on 11 typhoons, beats DSD-based parametric QPE for rainfall estimation in South China.
Observations & remote sensingChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)Nanjing University of Information Science & Technology (NUIST)
Convective Precipitation Nowcasting Using U-Net Model
U-Net on five radar frames predicts convective reflectivity, matching TrajGRU and outperforming TREC with a simpler CNN.
Using deep learning to emulate and accelerate a radiative-transfer model
Knowledge-guided U-Net++ emulates the RRTM shortwave radiation scheme for NWP, running about 10^4 times faster with good tropical generalization.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Forecasting excessive rainfall with random forests and a deterministic convection-allowing model
Random forests trained on convection-allowing WRF output probabilistically forecast excessive rainfall over CONUS, but they consistently underperform WPC's day-1 Excessive Rainfall Outlooks.
Use of Deep Learning for Weather Radar Nowcasting
Residual convolutional encoder-decoder models for radar echo nowcasting match or beat operational systems and transfer across regions without performance decay.
A Deep Learning Method for Bias Correction of ECMWF 24–240 h Forecasts
CU-net, a CNN, corrects gridded ECMWF-IFS 24–240 h forecasts of four surface variables over North China, beating the ANO correction at all lead times.
Post-processing & calibrationChina Meteorological Administration (CMA)Colorado State University (incl. CIRA)
CIRA Guide to Custom Loss Functions for Neural Networks in Environmental Sciences -- Version 1
A practical guide to writing custom loss functions for environmental and meteorological neural networks, including spatial verification scores like the fractions skill score and physical constraints.
Benchmarks, datasets & evaluationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Detecting Climate Signals Using Explainable AI With Single‐Forcing Large Ensembles
Layer-wise relevance propagation attributes regional temperature signals to aerosol versus greenhouse forcing in single-forcing large ensembles.
Climate analysis & attributionColorado State University (incl. CIRA)
Assessing Decadal Predictability in an Earth‐System Model Using Explainable Neural Networks
Explainable neural networks with layerwise relevance propagation identify North Atlantic and North Pacific SST regions that drive decadal predictability of continental temperature in an Earth-system model.
Climate analysis & attributionColorado State University (incl. CIRA)
Deep Learning Experiments for Tropical Cyclone Intensity Forecasts
An MLP using SHIPS predictors beats operational statistical-dynamical models by 5-22% for Atlantic tropical cyclone intensity in 2019-2020 real-time tests.
From Random Forests to Flood Forecasts: A Research to Operations Success Story
CSU-MLP random-forest excessive rainfall forecasts, built from ensemble reforecasts, were transitioned into operational use at NOAA's Weather Prediction Center.
Extremes & hazardsColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Subseasonal Forecasts of Opportunity Identified by an Explainable Neural Network
Neural network identifies subseasonal forecasts of opportunity over the North Atlantic at 22-day lead, with layer-wise relevance propagation revealing tropical predictors.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Machine Learning for Clouds and Climate (Invited Chapter for the AGU Geophysical Monograph Series "Clouds and Climate")
Review of interpretable and physics-guided machine learning for cloud processes in climate models, covering radiation, microphysics, convection and cloud detection, with a getting-started guide.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)Columbia University (LEAP, Lamont-Doherty)
Controlled abstention neural networks for identifying skillful predictions for classification problems
The NotWrong loss adds an abstention class so classifiers learn from confident samples, outperforming existing losses on several climate prediction tasks.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Controlled abstention neural networks for identifying skillful predictions for regression problems
An abstention loss lets regression networks flag low-confidence samples during training, outperforming a standard uncertainty baseline on synthetic climate tasks.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Applying machine learning methods to detect convection using Geostationary Operational Environmental Satellite-16 (GOES-16) advanced baseline imager (ABI) data
A convolutional network on one-minute GOES-16 visible and infrared imagery detects convection, trained against MRMS radar precipitation types as a proof of concept.
Observations & remote sensingColorado State University (incl. CIRA)
Neural Network Attribution Methods for Problems in Geoscience: A Novel Synthetic Benchmark Dataset
Synthetic attribution benchmark for XAI in generic geoscience regression problems; not specifically about weather or climate.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Random Forest Model to Assess Predictor Importance and Nowcast Severe Storms using High-Resolution Radar–GOES Satellite–Lightning Observations
Random forests combining GOES-16 satellite, lightning and radar fields improve severe storm warning prediction over radar alone, identifying cloud-top divergence as most informative.
NowcastingColorado State University (incl. CIRA)NASAUniversity of Oklahoma (AI2ES)
Predicting the MJO using interpretable machine-learning models
Interpretable neural networks forecast MJO convection several weeks ahead, using layer-wise relevance propagation to explain sources of predictability compared with dynamical models.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Advancing Radar Nowcasting Through Deep Transfer Learning
Transfers a CNN radar nowcaster trained near Beijing to Guangzhou via fine-tuning and MMD, improving skill with only small target-domain datasets.
Will Artificial Intelligence supersede Earth System and Climate Models?
Perspective proposing Neural Earth System Modelling, which fuses deep networks and Earth system models into interpretable hybrids that might eventually replace them.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)Potsdam Institute for Climate Impact Research (PIK)Princeton UniversityUniversity of Washington
ClimateNet: an expert-labeled open dataset and deep learning architecture for enabling high-precision analyses of extreme weather
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
Identifying Opportunities for Skillful Weather Prediction with Interpretable Neural Networks
Layer-wise relevance propagation applied to neural networks identifies atmospheric states associated with enhanced weather predictability, giving physical insight into predictability mechanisms.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Observational Constraints on Warm Cloud Microphysical Processes Using Machine Learning and Optimization Techniques
Machine-learning-constrained autoconversion and accretion parameterizations, fitted to Azores cloud-probe data, cut uncertainty to about 15% and 5% and reveal drizzle number's role.
Hybrid physics-ML & parameterizationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of ReadingUniversity of Washington
Assimilating GOES-R Latent Heating in FV3 using Machine Learning
A CNN called GREMLIN fuses GOES-R ABI and GLM data into synthetic radar reflectivity, proposed to feed convective-scale data assimilation and latent heating initialization.
Observations & remote sensingColorado State University (incl. CIRA)
Inferring Airmass Properties from GOES-R ABI Observations
CNNs infer airmass properties such as moisture and instability from GOES-R ABI imagery, offering an automated alternative to sparse radiosondes for NWP and forecasters.
Observations & remote sensingColorado State University (incl. CIRA)
AI in Weather Radars
Overview of AI for weather radar, covering quantitative precipitation estimation, hydrometeor classification and nowcasting using dual-polarization and spaceborne radar data.
Observations & remote sensingColorado State University (incl. CIRA)
Evaluation, Tuning, and Interpretation of Neural Networks for Working with Images in Meteorological Applications
Tutorial-style guidance on evaluating, tuning and interpreting neural networks for meteorological satellite images, covering receptive fields, relevance propagation and image-to-image translation.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Indicator patterns of forced change learned by an artificial neural network
A neural network trained to predict the year from climate model temperature and precipitation maps, with relevance visualization, reveals indicator patterns of forced climate change.
Climate analysis & attributionColorado State University (incl. CIRA)
Evaluation, Tuning and Interpretation of Neural Networks for Meteorological Applications
Practical guidance on evaluating, tuning and interpreting meteorological neural networks, covering effective receptive fields, relevant performance measures, and layer-wise relevance propagation for image classification and translation.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Forecasting Severe Weather with Random Forests
Random forests trained on GEFS reforecasts give calibrated Day 1-3 severe weather outlooks that beat SPC at Days 2-3, and blending with SPC improves Day 1 outlooks.
Development and Interpretation of a Neural Network-Based Synthetic Radar Reflectivity Estimator Using GOES-R Satellite Observations
A CNN converts GOES-R radiances and lightning into synthetic radar reflectivity, enabling convective radar assimilation with much better skill than pixel-wise methods.
Observations & remote sensingColorado State University (incl. CIRA)
Detection of Non‐Gaussian Behavior Using Machine Learning Techniques: A Case Study on the Lorenz 63 Model
Support vector machines and neural networks detect non-Gaussian error distributions in the Lorenz-63 model, informing relaxation of Gaussian assumptions in data assimilation.
The Importance of Past MJO Activity in Determining the Future State of the Midlatitude Circulation
Logistic regression with linear baroclinic model simulations shows the past 15 days of MJO activity improve midlatitude circulation predictability beyond 15 days ahead.
Sub-seasonal to seasonalColorado State University (incl. CIRA)
Physically Interpretable Neural Networks for the Geosciences: Applications to Earth System Variability
Applies backwards optimization and layerwise relevance propagation to neural networks, recovering scientifically meaningful patterns in climate variability data.
Explainability, uncertainty & physicsColorado State University (incl. CIRA)
Leveraging Modern Artificial Intelligence for Remote Sensing and NWP: Benefits and Challenges
Review arguing modern AI can augment or replace parts of satellite remote sensing and NWP processing chains, from observations to forecasts.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Viewing Forced Climate Patterns Through an AI Lens
A neural network trained to predict year from temperature maps detects forced climate patterns without explicitly separating signal from noise.
Climate analysis & attributionColorado State University (incl. CIRA)
Concluding the 2017 Hurricane Season: Evaluation of Impact Level Forecasts with Varied Meteorological Hazards
Evaluates a neural-network hurricane impact-level forecast over 2017 storms, finding best accuracy about 30 h before landfall and wind speed as the key driver.
Applications & impactsColorado State University (incl. CIRA)
A Machine Learning System for Precipitation Estimation Using Satellite and Ground Radar Network Observations
An MLP fuses geostationary and LEO satellite data with dual-polarization ground radar to improve precipitation estimates over Dallas-Fort Worth.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Using a Coevolutionary Postprocessor to Improve Skill for Both Forecasts of Surface Temperature and Nowcasts of Convection Occurrence
Coevolutionary evolutionary-programming postprocessor improves 72-h surface temperature forecasts and 60-min convection nowcasts over bias-corrected NWP and logistic regression.
Post-processing & calibrationColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Atmospheric Retrievals and Assessment for Microwave Observations from Chinese FY-3C Satellite during Hurricane Matthew
Neural network combined with 1D-variational retrieval of FY-3C microwave temperature and humidity profiles during Hurricane Matthew, improving WRF data-assimilation track forecasts.
Observations & remote sensingColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
Testing the Reliability of Interpretable Neural Networks in Geoscience Using the Madden-Julian Oscillation
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
Retrieving temperature and relative humidity profiles from hyperspectral radiations via deep learning
Retrieves temperature and humidity profiles from NPP CrIS hyperspectral radiances using autoencoders and a self-normalizing network trained on ECMWF data.
Observations & remote sensingColorado State University (incl. CIRA)
“Dendrology” in Numerical Weather Prediction: What Random Forests and Logistic Regression Tell Us about Forecasting Extreme Precipitation
Random forests and logistic regression trained on GEFS reforecast predictors forecast CONUS extreme precipitation at day 2 and day 3, and the paper analyzes what the trained models reveal physically.
Money Doesn't Grow on Trees, but Forecasts Do: Forecasting Extreme Precipitation with Random Forests
Random forests trained on 11 years of GEFS reforecasts give better CONUS extreme precipitation exceedance forecasts than raw ensembles and ECMWF.
Post-processing & calibrationColorado State University (incl. CIRA)
Causal Discovery for Climate Research Using Graphical Models
Accessible introduction to constraint-based causal discovery for climate scientists, illustrated on boreal winter atmospheric low-frequency modes.
Climate analysis & attributionColorado State University (incl. CIRA)