AI summary by TwinMind
How can climate be predictable if weather is chaotic?
Sabine Hossenfelder
7m 55s
Introduction and Core Question
- Common question: How can scientists predict climate 100 years ahead when weather forecasts fail beyond 2 weeks due to chaos?
- Standard answer "climate is not weather" is correct but insufficient explanation
- Goal: Explain how reliable long-term predictions possible despite impossible short-term predictions
Edward Lorenz and Discovery of Chaos (1963)
- Edward Lorenz: meteorologist who accidentally discovered weather is chaotic
- 1960s experiment: repeated calculation to predict weather trend
- Rounded initial value from 6 digits to 3 digits after decimal point
- Tiny difference in initial value → wildly different results
- Butterfly effect concept: butterfly flap in China might cause tornado in Texas 2 weeks later
Lorenz Model Explanation
- Lorenz simplified complex equations to 3 equations capturing strange behavior
- Three variables: X, Y, Z as functions of time (t)
- Interpretation: simplified description of convection in gases/fluids
- Can be integrated on laptop computer
Attractor Visualization
- Solution appears as curve in 3D space (X, Y, Z axes)
- Circles around two different locations back and forth
- All solutions end up doing circles near two central places = "attractor"
- Attractor shape resembles butterfly with two "wings"
Chaos Demonstration
- Two very similar (not identical) initial values
- Curves initially look similar, then diverge
- After some time become entirely uncorrelated
- Focus on X-direction coordinate for clarity
- Curves flip between positive/negative values (corresponding to attractor wings)
Weather Prediction Analogy
- Early range (up to t=5): decent weather forecast possible
- Beyond that: outcome depends sensitively on initial value
- Measurement error makes good prediction impossible after short period
- Analogy: strange planet weather flipping between no rain/pouring extremes
Average Calculations and Climate Analogy
- Calculate average rainfall over period (e.g., year)
- Integration over time period ÷ duration = average
- Curves from t=0 to t=100: average ≈ 0
- System spends equal time on each attractor wing
- Average depends on initial value, fluctuates around zero
- Longer averaging periods → closer to zero
Radiative Forcing Parameter
- Added constant "f" to model equations
- Mimics climate scientists' "radiative forcing"
- Radiative forcing = excess power per area Earth captures due to increasing CO2
- Relative to reference value
Forcing Effects Analysis
- f=1: looks similar to original
- f=2: minimal visible change
- f=3: system spending more time in upper range
- f=4: clearly spending more time in upper range
- f>7: system gets stuck on one attractor side immediately
- Effect occurs for all initial values despite chaos
Predictable Trends in Chaos
- Calculated average curves as function of radiative forcing
- Sample of initial values tested
- Clear correlation between average value and radiative forcing
- Scatter due to finite averaging period
- Key insight: trends of average values predictable even when exact states unpredictable beyond short periods
Climate Model Application
- Scientists cannot predict rain on specific date (June 15, 2079)
- Can predict average rainfall in 2079 as function of increasing CO2 levels
- Demonstrates how long-term climate trends predictable despite chaotic short-term weather
