ATM S 111, Global Warming: Understanding the Forecast

1. ATM S 111, Global Warming: Understanding the Forecast Dargan M. W. Frierson Department of Atmospheric Sciences Day 15: 05/18/2010

2. Climate Models

3. Next Topic: Climate Models • Predicting the climate using computers • Important tool for understanding possible future climates • Also the topic of most of my research!

4. Using Climate Models to Build Understanding • Often climate models are thought of as forecast tools (what’s the climate going to be like in 50 years?) • Models are equally useful for developing understanding though • We only have one Earth to observe • We’re only limited by our creativity in making our own computer worlds

5. Climate Models • We’ll discuss via : • The discovery of chaos by Ed Lorenz • The first climate models of SukiManabe • But first: • Climate models are closely related to weather prediction models • Let’s discuss some history of weather prediction using computer models • And the first attempt at numerical weather forecasting by Lewis Fry Richardson

6. Numerical Weather Prediction (NWP) Improvements in weather prediction over the last 60 years are among the most impressive accomplishments of society Northern Hem. 3 day forecast Southern Hem. 3 day forecast 5 day forecasts (NH & SH) 7 day forecasts (NH & SH) 2002 1980 Simmons & Hollingsworth ‘02

7. Lewis Fry Richardson British mathematician, physicist, atmospheric scientist Scientific career very influenced by his Quaker beliefs (pacifism) Made the first numerical weather prediction in 1922 Also had a dream of the future of weather prediction…

8. Richardson’s Dream: The Forecast Factory Filled with employees (“computers”) doing calculations Richardson’s dream in 1922 of a global forecasting system He estimated 64,000 “computers” (people) would be necessary to forecast over the globe Much info from the next few slides is from a book by Peter Lynch (U Coll Dublin)

9. Richardson’s Experiment Used data from May 20, 1910 “Leipzig charts” Surface pressure and temperature

10. Richardson’s Experiment Data taken when Halley’s Comet was passing through the atmosphere Tabulated values from these charts by hand! Upper atmosphere temperature and pressure

11. Richardson’s Calculations • Served as ambulance driver with the Friends’ Ambulance Unit in France during WWI • Transported injured soldiers, often under heavy fire • Took 1000 hours of work to perform the calculations • “My office was a heap of hay in a cold rest billet” • Calculation book was lost during the battle of Champagne • But recovered months later under a heap of coal • Eventually published in 1922

12. Failure or Success? • First prediction was for pressure to change by 145 mbar in 6 hours • Hugely, hugely wrong…. • Richardson himself realized that noisy wind data was likely the problem • He suggested 5 different filtering methods to fix this • Obviously he couldn’t try this experiment again • But we can reproduce the results using today’s computers…

13. Computer Forecast w/ Richardson’s Proposed Fix A good forecast!

14. First Weather Prediction on Computers • 1946: test case by von Neumann and others • May 1955: Joint Numerical Weather Prediction Unit, Maryland • First operational computer forecasts in US • Global coverage since 1973 • Computers surpassed human forecasts: 1980s?

15. Weather Forecasting vs Climate Forecasting • Weather models and climate models are similar in a lot of ways • Use very similar mathematical equations • But weather forecasting and climate forecasting have very different goals • How can we predict the climate in 50 years if we can’t predict the weather 2 weeks from now?

16. Chaos • Ed Lorenz was running a computer model & put in slightly different inputs • He found the predictions were similar for a while but then wildly diverged to different solutions • Chaos: when small changes make a big & unpredictable difference Edward Lorenz (1908 - 2008) meteorologist, M.I.T. father of chaos theory

17. Limit to weather forecast skill: Chaos "Does the flap of a butterfly's wings in Brazil set off a tornado in Texas?" [Lorenz, 1972] Weather forecasts depend very sensitively on the initial observations. We can’t observe every butterfly flapping its wings, weather forecasts lose “skill” (ability to predict storms, not just the right season) after ~2 weeks. In contrast, climate models are all about modeling seasons… Edward Lorenz (1908 - 2008) meteorologist, M.I.T. father of chaos theory see Rough Guide, p. 228

18. Climate Forecasts • This limit to weather prediction doesn’t affect climate forecasts • It all averages out after a month or so of storms • Climate forecasts: • Summer is hotter than winter • After a strong volcano blows up, the Earth will cool • The Earth will be hotter with more greenhouse gases • Shifts in weather patterns when El Niño is present • Etc…

19. SukiManabe: Father of Climate Modeling • SyukuroManabe (born 1931): • Worked at GFDL from 1958-1997 • 1997-2001: Director of Earth Simulator, Japan My grand-advisor (via Isaac Held)

20. Pioneer in Climate Modeling ”Let the model tell you the answer" paraphrased from Manabe, 2005 Make the model physics as accurate as you possibly can and it will be better at predicting the climate than you could reason with pencil and paper. (an attempt to describe his philosophy) Syukuro Manabe meteorologist, Princeton

21. Early Manabe Modeling Studies • Gradually building up more sophisticated climate models: • Radiation only model (longwave and shortwave): M. and Moller (1961) • Above plus convection: M. and Strickler (1964) • Model with atmospheric motions (but no ocean yet): Smagorinsky, M. and Holloway (1965)

22. First Coupled Climate Model • Manabe and Bryan (1969): • First coupled climate model

23. First Global Warming Forecast Manabe and Wetherald (1975): Polar amplification Wet areas get wetter & subtropical drying

24. Other Early Manabe Studies • I find these early modeling papers still really fascinating… • Effect of ocean circulation on climate: • Turn off ocean model • Effect of moisture: • Don’t allow condensation to occur • Effect of mountains: • Bulldoze all topography • Effect of changing solar radiation, doubling CO2, ice sheets, clouds, soil moisture, etc…

25. GCM Components • GCM: Global Climate Model • Components of GCMs: • Equations of fluid motion on a rotating sphere • Both the atmosphere and the ocean are just fluids • Equations put simple physics principles in mathematical form: • Mass is neither created nor destroyed • Heating/cooling changes temperature • Forces & pressure change momentum • Compressibility of air • Etc etc

26. Dynamical Core of AGCMs Essentially just fluid equations on the rotating sphere

27. Climate Models Chop up the Earth into Grid Cells

28. A close up of Europe • The current horizontal size of an atmosphere, land, ocean or sea ice grid cell is about 150km x 150km The vertical extent of a box is typically: Atmosphere/Ocean: 80-500m Sea Ice: 50cm Land: 10cm

29. Model Resolution Evolution Changes in resolution over time: (1990) (2001) AR = “assessment report” of IPCC FAR = “first” IPCC assesment report, etc. (1995) (2007)

30. Model Resolutions

31. GCM Components • GCM: Global Climate Model • Components of GCMs: • Equations of fluid motion on a rotating sphere • Heat sources • Shortwave and longwave radiation • Condensation • Surface fluxes • Have to parameterize small-scale processes • Clouds • Moist convection

32. Within each grid cell, there are things that are not explicitly modeled (e.g., clouds) that must be approximated or “parameterized” (e.g, the fraction of clouds may be an empirical function of temperature and humidity)

33. Cloud schemes • Cloud interactions are the most uncertain process in GCMs • Lead to the largest differences between models

35. Weather models have cool looking visualizations of their output

36. While we often look as something as simple as global mean temperature from climate models (no one would EVER look at global mean temperature in a weather model) 2007 IPCC Figure

37. Average Northwest warming, 2000-2100 Degrees C Grey shading includes predictions from 10 models for two scenarios. Why is the range larger locally than globally? Climate variability like the Pacific Decadal Oscillation can lead to cooling of some regions but warming of others Source: Mote, Salathé and Peacock 2005

38. CCSM3 Arctic September Sea Ice Projections 7 runs from one climate model (no one would EVER look at sea ice in a weather model) red = observations colors = 7 runs black = mean The initial conditions are varied here on purpose. Why?

39. Surface Temperature Change at end of 21st century relative to end of last century

40. Climate models have weather variability, they just aren’t designed to make day-to-day forecasts

41. Climate modelers use cooler computers With a peak speed of 2.33 petaflops (over two thousand trillion calculations per second), "Jaguar," a Cray XT5 supercomputer. Owned by the Department of Energy. Not just for climate science. But not used for weather.

42. Highest resolution models can capture more details of cloud structures This will be the resolution of the future

43. How do we know if climate models are right?

44. Annual Average Surface Temperature Observed Model Average ºC IPCC 2007

45. Annual Average Surface Temperature Error in a typical model IPCC 2007

46. “Annual Cycle*” in Temperature * Multiply by ~3 to get approximately the difference in July and January temperature Observed Model Average IPCC 2007

47. Annual Average Precipitation Observed (cm/year) Average of the models IPCC 2007

48. Other Ways to Validate Climate Models How much cooling after a volcano? Can we reproduce the last Ice Age conditions given CO2, solar, etc conditions? Can the climate of the 20th century be reproduced given greenhouse gas, solar, volcanoes, and aerosols?

49. “Prediction is very difficult, especially about the future” Niels Bohr Niels Bohr with Albert Einstein

50. Climate model projection made in 1980: How well did it do? Models: heavily smoothed Observations: 5-year running mean Reference period: 1961-1990 In 1980, little was known about how fast CO2 would rise. Version C was a more modest assumption.