Using reinforcement learning to plan for an uncertain climate future

Science Sessions are brief conversations with cutting-edge researchers, National Academy members, and policymakers as they discuss topics relevant to today's scientific community. Learn the behind-the-scenes story of work published in the Proceedings of the National Academy of Sciences (PNAS), plus a broad range of scientific news about discoveries that affect the world around us.

In this episode, Ning Lin talks about how reinforcement learning methods plant to mitigate climate risk despite uncertainty in climate change risk forecasts.

In this episode, we cover: •[00:00] Introduction •[1:04] Civil engineer Ning Lin introduces why climate forecast uncertainty complicates risk management planning. •[02:41] Lin explains how reinforcement learning works. •[03:26] She talks about why the team studied risk management for Manhattan. •[04:54] Lin explains the results of the reinforcement learning study. •[05:40] She recounts the results that surprised her. •[07:25] Lin explains the takeaways from the study for emergency planners. •[09:00] She enumerates the caveats and limitations of the study. •[10:11] Conclusion.

About Our Guest:

Ning Lin Professor Princeton University

View related content here: https://www.pnas.org/cgi/doi/10.1073/pnas.2402826122

Follow us on Spotify, Apple Podcasts, or wherever you get your podcasts for more captivating discussions on scientific breakthroughs!

Visit Science Sessions on PNAS.org: https://www.pnas.org/about/science-sessions-podcast

Follow PNAS: Twitter/X Facebook LinkedIn YouTube Sign up for the PNAS Highlights newsletter