AI summary by TwinMind
Reality is worse than climate models predict, new study finds | Science News
Sabine Hossenfelder
5m 55s
Origins of Event Attribution
- Miles Allen (Oxford University) first proposed event attribution concept in Nature commentary 2003
- Original motivation: determine if fossil fuel industry could be sued for flooding
- Oxford flooded again last week, no successful lawsuits yet
- Student Fred Otto (now Professor) developed concept further
- Collaborated with Gan F Orden Bor (deceased)
- Recognized numbers not useful for lawsuits but valuable for media headlines
Event Attribution Methodology
- Simple process: specify extreme event type (e.g., flood above certain rainfall threshold)
- Use climate model to make two predictions for event frequency:
- Current greenhouse gas levels
- Pre-industrial levels
- Compare two cases to determine climate change impact
- Results often presented as "X times more likely" due to climate change
Multiple Problems with Current Approach
- Event frequency depends strongly on exact event definition
- More details added = less likely event appears in simulations
- Can result in comparing zero to zero (meaningless)
- Definition manipulation possible to achieve desired results
- Climate models not designed to simulate extreme weather events
- Not effective at predicting extreme events
- Most models underestimate both frequency and severity of extreme weather
Scientific Evidence of Model Limitations
- 2021 review paper findings:
- "Climate models have not been designed to represent extremes well"
- Trends differ widely between different climate models
- Attribution study results depend strongly on quantitative event definition
- New paper examining 21 best climate models vs observed changes:
- "Virtually all climate model significantly underestimate the rates at which increases in precipitation extremes have scaled with global temperatures historically"
Policy Implications and Misinterpretation
- Event attribution creates false sense of accuracy
- Example: "twice as likely" flood misleads policymakers
- They assume 50 years until next occurrence
- Reality: could happen annually or twice yearly
- Numbers represent minimum increase in likelihood, not precise predictions
Proposed Solution
- Entire event attribution field should be regarded with great skepticism
- Proposes new research area: "attribution attribution"
- Easier to attribute popularity of attribution to climate change than actual events
