Rapid Radiative Transfer Model for General Circulation Models (RRTMG)

Gold definitionUpdated Apr 2, 2026

Definition

RRTMG is a computationally intensive component in General Circulation Models that simulates atmospheric radiation. Machine learning is increasingly used to approximate RRTMG to significantly improve the computational efficiency of numerical weather prediction and climate models.

At a glance

Executive summary

RRTMG is a key part of weather and climate models that calculates how radiation moves through the atmosphere, but it's very slow. Researchers are now using AI, like neural networks, to speed up these calculations significantly, making weather predictions faster and more efficient.

TL;DR

RRTMG is a slow but important radiation calculator in big weather models, and AI is being used to make it much faster.

Key points

Simulates atmospheric radiative transfer within General Circulation Models (GCMs).
Addresses the computational bottleneck of radiation calculations in numerical weather and climate models.
Used by global operational weather and climate forecasting centers, e.g., China Meteorological Administration.
ML approximation offers significant speedup compared to traditional, computationally intensive physical parameterizations.
Research trend involves integrating deep neural networks into numerical prediction models for physical process emulation, focusing on stability and compatibility.

Use cases

Accelerating global operational weather forecasting systems by replacing traditional RRTMG with ML surrogates.
Enabling higher spatial and temporal resolution in climate simulations due to reduced computational cost of radiation.
Facilitating quicker experimentation and sensitivity studies in atmospheric science by speeding up model runs.
Deploying advanced numerical models on less powerful computing infrastructure by offloading intensive calculations to efficient ML models.

Also known as

RRTMG, Rapid Radiative Transfer Model, ML-RRTMG, Neural Network RRTMG