Quantifying “climate distinguishability” after stratospheric aerosol injection using explainable artificial intelligence

[Geoengineering News]

https://essopenarchive.org/users/622124/articles/645378-quantifying-climate-distinguishability-after-stratospheric-aerosol-injection-using-explainable-artificial-intelligence

Authors

• Antonios Mamalakis
• Elizabeth A. Barnes
• James Wilson Hurrell

Peer review timeline

25 May 2023Submitted to ESS Open Archive 

01 Jun 2023Published in ESS Open Archive

Cite as: Antonios Mamalakis, Elizabeth A. Barnes, James Wilson Hurrell. Quantifying “climate distinguishability” after stratospheric aerosol injection using explainable artificial intelligence. ESS Open Archive . June 01, 2023.
DOI: 10.22541/essoar.168565419.96137434/v1

Abstract

Stratospheric aerosol injection (SAI) has been proposed as a possible complementary solution to limit global warming and its societal consequences. However, the climate impacts of such intervention remain unclear. Here, we introduce an explainable artificial intelligence (XAI) framework to quantify how distinguishable an SAI climate might be from a pre-deployment climate. A suite of neural networks is trained on Earth system model data to learn to distinguish between pre- and post-deployment periods across a variety of climate variables. The network accuracy is analogous to the “climate distinguishability” between the periods, and the corresponding distinctive patterns are identified using XAI methods to gain insights into the emerging signals from SAI. For many variables, the two periods are less distinguishable under SAI than under a no-SAI scenario, suggesting that the specific intervention modeled decelerates future climatic changes. Other climate variables for which the intervention has negligible effect are also highlighted.

Source: ESS OPEN ARCHIVE