Probabilistic Emulation of the Community Radiative Transfer Model Using Machine Learning
The episode discusses a paper creating a machine learning emulator for the Community Radiative Transfer Model (CRTM) to speed up satellite data processing. The team developed a neural network that runs about five times faster than CRTM and provides probabilistic error estimates, allowing forecasters to use the emulator when confident and fall back on CRTM when uncertain. This approach aims to increase the amount of satellite data used in weather forecasts.