Day 1: Bjorn Stevens - Three (Plus One) Problems of Climate Science

Value of the Information & Strength of the Argument

The talk provides valuable insights into the current state and future direction of climate modeling. Stevens clearly articulates the limitations of traditional approaches and makes a compelling case for km-scale simulations. His argument is well-structured, moving from fundamental questions of predictability to concrete computational examples. He supports his claims with specific examples, such as the linear relationship between cumulative CO2 emissions and temperature, and the spread in model projections due to cloud feedback. The discussion of structural stability and the convergence of models at high resolution is particularly thought-provoking. However, some statements are forward-looking and not yet fully demonstrated, such as the claim that models will converge to the same answer by the end of the decade. Overall, the argumentation is solid and based on established physics, but it is a keynote address rather than a rigorous scientific paper.

Scientific Rigor, Source Quality, Title Accuracy

The speaker is a highly credible expert, and the talk reflects deep knowledge of the field. He references his own work and collaborations, such as the scaling analysis with Torson Hler, and mentions specific supercomputers (Jupiter) and model results. However, the talk does not provide detailed citations or references to specific publications, which limits the ability to verify all claims. The title accurately reflects the content, and the talk stays on topic. There is no obvious bias or unsupported speculation, but the forward-looking statements about future capabilities should be taken as expert opinion rather than established fact.

The title accurately reflects the content: the speaker discusses three core problems (predictability, cloud response, regional impacts) plus a fourth (computational feasibility) in climate science.

8/10

The speaker is a leading climate scientist (Director of Max Planck Institute for Meteorology). The talk presents a well-reasoned argument based on established physics and recent computational results, but it is a keynote presentation rather than a peer-reviewed study. Some claims about model convergence and future capabilities are forward-looking and not yet fully verified.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 01:00Introduction and visualization of weather chaos and climate attractor.
• 03:00Discussion of the predictability problem and the lack of theory for climate predictability.
• 07:00Explanation of traditional climate models and their limitations.
• 10:00Presentation of the linear relationship between CO2 emissions and temperature, and the spread due to cloud feedback.
• 13:00Discussion of regional climate changes and the challenge of defining what 2°C means locally.
• 16:00Introduction of the new approach: computing climate from weather using km-scale simulations.
• 19:00Scaling analysis of computational throughput and the Jupiter supercomputer.
• 22:00Results showing convergence of models at 1 km resolution.
• 25:00Discussion of structural stability and the need for multiple models.
• 28:00Vision for simulating at human scale and creating databases for imagining future climates.

• ADIA Lab Symposium 2025 — Event page for the symposium where this talk was given.

• IPCC Sixth Assessment Report — The IPCC report discusses climate projections and uncertainties, consistent with the talk's themes.