paper
2023
Hybrid physics-ML & parameterization
New model or method
CNN / U-NetColorado State University (incl. CIRA)NOAA (incl. GFDL, PSL, GSL, NCEP/EMC) Journal
2023-08-31
055
Aug 2023
Ryan Lagerquist, David D. Turner, Imme Ebert‐Uphoff et al.
Journal of Atmospheric and Oceanic Technology
Colorado State University (incl. CIRA), NOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
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 & parameterization
Estimating Full Longwave and Shortwave Radiative Transfer with Neural Networks of Varying Complexity
In one sentence. 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.
Labels
- Topic
- Hybrid physics-ML & parameterization
- Type
- New model or method
- Method
- CNN / U-Net
- Domain
- Global
- Data
- NWP analysis or forecasts (IFS, GFS, ...), ERA5 / reanalysis
- Status
- Pre-operational / experimental
- Identifiers
- DOI:10.1175/jtech-d-23-0012.1 OpenAlex:W4386325634
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.