★ Regional data-driven weather modeling with a global stretched-grid
Graph-based stretched-grid model refines a global AI model to 2.5 km over the Nordics, beating the MEPS ensemble mean for 2 m temperature but underestimating extremes.
Regional & km-scale models
Limited-area and regional ML weather models, convection-permitting and km-scale emulators, including boundary handling and stretched grids.
Search and filter this topic Plain list on GitHub
★ Highlights
★ Kilometer-Scale Convection Allowing Model Emulation using Generative Diffusion Modeling
StormCast, a generative diffusion emulator of the 3 km HRRR convection-allowing model, gives competitive 1-6 hour radar reflectivity skill with realistic cold pools and power spectra.
All papers (17)
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.
Colorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)University of Oklahoma (AI2ES)
Building Machine Learning Limited Area Models: Kilometer-Scale Weather Forecasting in Realistic Settings
Kilometre-scale graph-based limited-area models for Denmark and Switzerland, driven by flexible boundary forcing, match or beat the NWP baseline for key surface variables.
Diffusion-LAM: Probabilistic Limited Area Weather Forecasting with Diffusion
Diffusion-LAM is a conditional diffusion model that generates probabilistic limited-area forecasts from boundary conditions, demonstrated on the MEPS regional ensemble dataset.
★ Regional data-driven weather modeling with a global stretched-grid
Graph-based stretched-grid model refines a global AI model to 2.5 km over the Nordics, beating the MEPS ensemble mean for 2 m temperature but underestimating extremes.
★ Kilometer-Scale Convection Allowing Model Emulation using Generative Diffusion Modeling
StormCast, a generative diffusion emulator of the 3 km HRRR convection-allowing model, gives competitive 1-6 hour radar reflectivity skill with realistic cold pools and power spectra.
MetMamba: Regional Weather Forecasting with Spatial-Temporal Mamba Model
MetMamba uses a Mamba state-space backbone for regional forecasting, outperforming attention and neural operator backbones and enabling coupled limited-area training with a global host.
YingLong-weather: AI-Based Limited Area Models for Forecasting of Non-precipitation Surface Meteorological Variables
YingLong is a 3 km AI limited-area model that beats WRF-ARW on surface wind speed in two domains while running much faster, using Pangu-Weather boundaries.
A comparison of data-driven reduced order models for the simulation of mesoscale atmospheric flow
Compares data-driven reduced-order models for simulating mesoscale atmospheric flow; the abstract is missing, so the ML contribution is uncertain.
Graph-based Neural Weather Prediction for Limited Area Modeling
Adapts graph-based neural weather prediction to limited-area modelling with a multi-scale hierarchical extension, validated on a Nordic regional model.
Data Driven Regional Weather Forecasting: Example using the Shallow Water Equations
Builds data-driven regional forecasting rules from local observations of shallow-water flow, predicting observed variables without governing equations.
Data Driven Regional Weather Forecasting
Data-driven forecasting rules built from regional observations predict shallow-water flow on a beta plane without explicit physical equations.
A case study of spatiotemporal forecasting techniques for weather forecasting
Comparison of spatiotemporal data-driven methods for regional gridded weather forecasting; tensor train DMD matches state-of-the-art accuracy without training.
Short-term precipitation prediction using deep learning
A 3D CNN from one meteorology frame predicts daily CONUS precipitation, beating operational weather models up to 5-day leads and improving them when combined.
Deep Learning Based Forecasting of Indian Summer Monsoon Rainfall
ConvLSTM precipitation forecasts over India at 25 km reach correlation 0.67 at lead day 1 and 0.42 at lead day 2.
Machine Learning for Climate Precipitation Prediction Modeling over South America
Machine learning precipitation forecasts over South America reduce errors to under 2 mm versus INPE's operational model, outperforming it in some regions.
Deep coastal sea elements forecasting using U-Net based models
U-Net variants with residual, parallel and asymmetric convolutions forecast multi-step coastal Netherlands sea elements from hourly Copernicus data, with the best model winning on frame prediction.
STConvS2S: Spatiotemporal Convolutional Sequence to Sequence Network for Weather Forecasting
STConvS2S uses only convolutional layers for sequence-to-sequence weather forecasting of South American temperature and rainfall, 23% better than an RNN baseline and five times faster to train.
Also relevant (23)
Generative Data Assimilation of Sparse Weather Station Observations at Kilometer Scales
Score-based diffusion data assimilation at km scale assimilates 40 sparse weather stations, lowering held-out station RMSE by 10% versus HRRR.
★ FengWu-GHR: Learning the Kilometer-scale Medium-range Global Weather Forecasting
First data-driven global model at 0.09 degree resolution, initialized by transfer from a low-resolution model, outperforming IFS-HRES in 2022 hindcasts.
Global medium-range NWPChina Meteorological Administration (CMA)Nanjing University of Information Science & Technology (NUIST)Shanghai Artificial Intelligence Laboratory
Post-processing using deep learning to create operational, high-resolution, and probabilistic weather forecasts
Operational multi-task neural network post-processes HRRR forecasts with a CRPS loss, improving RMSE by 3–25% and CRPS by 31–45% over HRRR.
Enhancing wind field resolution in complex terrain through a knowledge-driven machine learning approach
ESRGAN-inspired network upscales coarse wind fields to high-resolution 3D flow over complex terrain at a Norwegian wind farm, outperforming trilinear interpolation.
DSRSS-Net: Improved-Resolution Snow Cover Mapping from FY-4A Satellite Images Using the Dual-Branch Super-Resolution Semantic Segmentation Network
DSRSS-Net jointly super-resolves and segments FY-4A imagery for snow-cover mapping on the Tibetan Plateau, improving classification accuracy by 4.45% over MOD10A1.
Observations & remote sensingNanjing University of Information Science & Technology (NUIST)China Meteorological Administration (CMA)
A high-precision prediction method for coarse grids based on deep learning and the Weather Research and Forecasting model
Deep learning method improves WRF predictions on coarse grids; the abstract is unavailable, so the approach is inferred from the title only.
Using CNN framework to improve multi-GCM ensemble predictions of monthly precipitation at local areas: An application over China and comparison with other methods
CNN combines multi-GCM ensembles to improve monthly precipitation predictions at local scales over China compared with other methods.
Post-processing & calibrationChina Meteorological Administration (CMA)
Research on Radar Echo Extrapolation Method by Fusing Environment Grid Point Field Information
A 3D CNN that fuses environmental grid fields with radar echoes raises CSI by 3.42% at 40 dBZ for convective extrapolation.
Impact of Horizontal Resolution on the Robustness of Radiation Emulators in a Numerical Weather Prediction Model
Tests a machine-learning radiation emulator in an NWP model across 0.25–100 km resolutions over Korea, finding larger heating-rate errors at fine scales.
Hybrid physics-ML & parameterizationKorea Meteorological Administration (KMA)
Improving regional climate simulations based on a hybrid data assimilation and machine learning method
Couples data assimilation with ML to optimise WRF land-surface states in the Heihe basin, improving temperature and humidity simulations and ET estimates.
Downscaling of Historical Wind Fields over Switzerland Using Generative Adversarial Networks
GAN with gradient-penalized Wasserstein loss downscales ERA5 hourly winds over Switzerland from 25 km to 1.1 km, capturing orographic effects missed by ERA5.
Using machine learning to produce a cost-effective national building height map of Ireland to categorise local climate zones
Builds the first open 10 m building height map of Ireland from Sentinel-1/2 features; CNN is most accurate, but ridge regression is chosen for realism and low cost.
Wind‐Topo: Downscaling near‐surface wind fields to high‐resolution topography in highly complex terrain with deep learning
Wind-Topo uses deep learning to downscale COSMO-1 near-surface winds to 50 m over Swiss Alpine topography, reproducing orographic effects at 60 validation stations.
Use of artificial neural network for generation of ultra-high resolution meteorological fields over Chowkibal-Tangdhar region of North-West Himalaya
Neural networks post-process 2 km WRF output over the NW Himalaya and interpolate it to 90 m, with NSE above 0.75 at five stations.
Combining computational fluid dynamics and neural networks to characterize microclimate extremes: Learning the complex interactions between meso-climate and urban morphology
Trains feedforward and deep neural networks on CFD simulations to downscale mesoscale wind and temperature to urban microscale, reaching R2 of 0.91 and 0.94.
Downscaling & super-resolutionColumbia University (LEAP, Lamont-Doherty)
Detection of Bow Echoes in Kilometer-Scale Forecasts Using a Convolutional Neural Network
A CNN detects bow echoes in French kilometre-scale model output with 86% hit rate and 39% false alarms, and forecasters found the products useful.
Addressing up-scaling methodologies for convection-permitting EPSs using statistical and machine learning tools
Two machine-learning and statistical up-scaling methods for convection-permitting ensemble rainfall forecasts reduce the double-penalty problem and improve severe-weather forecast skill.
Modeling high-resolution precipitation by coupling a regional climate model with a machine learning model: an application to Sai Gon–Dong Nai Rivers Basin in Vietnam
Couples a regional climate model with a machine learning model to produce high-resolution precipitation for the Sai Gon–Dong Nai basin in Vietnam; abstract unavailable, classified from title.
Statistical Post-Processing for Gridded Temperature Prediction Using Encoder-Decoder-Based Deep Convolutional Neural Networks
An encoder-decoder CNN post-processes NWP output to produce gridded surface temperatures over Kanto, correcting front position errors and extreme temperatures better than the operational Kalman-filter guidance.
Post-processing & calibrationJapan Meteorological Agency (JMA)
Valley Winds at the Local Scale: Correcting Routine Weather Forecast Using Artificial Neural Networks
A neural network corrects 3 km WRF forecasts of valley winds in complex terrain, significantly improving low-level wind speed and direction.
Fast and accurate learned multiresolution dynamical downscaling for precipitation
CNNs trained with MSE and conditional GANs emulate 12 km precipitation from 50 km simulations, running 30 years in 14 minutes versus one month.
Deep Learning for Climate Model Output Statistics
ConvMOS, a CNN for model output statistics, reduces precipitation errors in the REMO regional climate model and outperforms three common MOS approaches.
Super-Resolution Simulation for Real-Time Prediction of Urban Micrometeorology
CNN super-resolution applied to low-resolution urban micrometeorology simulations gives 4x finer predictions faster than high-resolution runs, beating interpolation.