Machine-learning in radiative transfer

J-.C. Thelen

doi:10.1117/12.2618774

27 May 2022 Machine-learning in radiative transfer

J-.C. Thelen

Proceedings Volume 12109, Thermosense: Thermal Infrared Applications XLIV; 121090H (2022) https://doi.org/10.1117/12.2618774
Event: SPIE Defense + Commercial Sensing, 2022, Orlando, Florida, United States

Abstract

Radiative transfer (RT) codes have many applications ranging from weather/climate predictions and atmospheric sciences to remote sensing and astrophysics. However, traditional RT codes are computationally very expensive and increasingly unable to process the large amounts of data resulting from modern global circulation models (GCM) or satellite feeds. One way to alleviate this problem is to use statistical emulators, i.e., fast and accurate approximate models based on statistical inference, to replace the deterministic RT codes. In his paper, we develop a statistical surrogate model which allows us to predict the radiances or brightness temperatures, i.e., the amount of electromagnetic energy measured by an electro-optical sensor, from the atmospheric state variables. The emulator is based on Gaussian Processes (GPs) which, for our purposes, are deemed to have several advantages over neural networks (NN). Unlike neural networks, GPs provide an analytical expression for the predictive error, the underlying model is interpretable and differentiable and the datasets required for training the model are considerably smaller than those for NNs.

Conference Presentation

Citation Download Citation

J-.C. Thelen "Machine-learning in radiative transfer", Proc. SPIE 12109, Thermosense: Thermal Infrared Applications XLIV, 121090H (27 May 2022); https://doi.org/10.1117/12.2618774

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available