ORBIT-2: Scaling Exascale Vision Foundation Models for Weather and Climate Downscaling
ORBIT-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.
Oak Ridge National Laboratory
17 papers in the archive.
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
ORBIT: Oak Ridge Base Foundation Model for Earth System Predictability
ORBIT scales a vision transformer to 113 billion parameters for Earth system prediction, sustaining 1.6 exaFLOPS on 49,152 GPUs of the Frontier supercomputer.
Assessing Tropical Pacific–Induced Predictability of Southern California Precipitation Using a Novel Multi-Input Multioutput Autoencoder
Multi-input autoencoder index from tropical Pacific SST gives Southern California precipitation predictability up to 4 months, beating Niño-3.4.
A Review of Recent and Emerging Machine Learning Applications for Climate Variability and Weather Phenomena
Review of ML for climate variability modes, extreme weather precursors, observation-model integration, downscaling and bias correction, aimed at domain scientists.
Climate analysis & attributionNSF NCAR / UCARUniversity of California, BerkeleyLawrence Berkeley National Laboratory / NERSCOak Ridge National Laboratory
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 & parameterizationOak Ridge National LaboratoryColorado State University (incl. CIRA)
A Spatiotemporal-Aware Climate Model Ensembling Method for Improving Precipitation Predictability
Bayesian neural network weights CMIP6 models per location and season, improving monthly US precipitation predictions over simple ensemble baselines with interpretable weights.
Multimodel Ensemble Predictions of Precipitation using Bayesian Neural Networks
Bayesian neural network learns spatiotemporally varying weights to combine climate-model precipitation ensembles, outperforming three ensemble methods and quantifying epistemic uncertainty.
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 & hazardsOak Ridge National LaboratoryColorado State University (incl. CIRA)
Reconstructing High Resolution ESM Data Through a Novel Fast Super Resolution Convolutional Neural Network (FSRCNN)
FSRCNN-ESM super-resolves coarse Earth system model output from 1° to 0.25° over North America, outperforming other super-resolution methods on temperature, fluxes and precipitation.
Super Resolution Reconstruction of E3SM Data Using a FSRCNN
FSRCNN-ESM, a modified fast super-resolution CNN, downscales E3SM climate output from 1° to 0.25° over North America with better accuracy than baselines.
A SPATIOTEMPORAL-AWARE WEIGHTING SCHEME FOR IMPROVING CLIMATE MODEL ENSEMBLE PREDICTIONS
A Bayesian neural network ensemble weights CMIP6 models by location and season, calibrating projections to observations with interpretable weights and uncertainty growing away from constrained periods.
Deep Learning Emulation of Subgrid‐Scale Processes in Turbulent Shear Flows
Deep neural networks emulate subgrid viscosity and diffusivity for turbulent stratified shear flows, giving a 2-2.5x simulation speedup while preserving kinetic and density variance budgets.
Hybrid physics-ML & parameterizationOak Ridge National Laboratory
A Causal Inference Model Based on Random Forests to Identify the Effect of Soil Moisture on Precipitation
Random-forest-based Granger causal framework isolates soil moisture to precipitation feedback over the US and uses it to benchmark Earth system models for land-atmosphere coupling.
Using Deep Neural Networks as Cost‐Effective Surrogate Models for Super‐Parameterized E3SM Radiative Transfer
Deep neural networks emulate SP-E3SM shortwave and longwave radiation with 90–95% accuracy at 8–10 times lower cost than the original scheme.
Hybrid physics-ML & parameterizationOak Ridge National Laboratory
Efficient surrogate modeling methods for large-scale Earth system models based on machine learning techniques
SVD plus a Bayesian-optimized neural network emulates 42,660 carbon-flux outputs of an Earth system ecosystem model from only 20 simulations.
Exascale Deep Learning for Climate Analytics
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