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 : But first:
The discovery of chaos by Ed Lorenz
The first climate models of Suki Manabe
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)
1980
2002
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
( )
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
( )
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. Suki Manabe: Father of Climate Modeling
Syukuro Manabe (born 1931):
Worked at GFDL from
: Director of Earth Simulator, Japan
My grand-advisor (via Isaac Held)

20. meteorologist, Princeton
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: m 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

34. Relevance of initial conditions
Weather Model
Climate Model
Goal
Predict weather
Predict climate
Time Range
days
years
Spatial Resolution
5-10 km
20-100km
Relevance of initial conditions
high (thus taken from weather balloon network)
low (only the ocean and sea ice matter much)
Relevance of GHG concentration
low
high
Relevance of ocean dynamics
Relevance of energy balance

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. (no one would EVER look at sea ice in a weather model)
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. Error in a typical model
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:
In 1980, little was known about how fast CO2 would rise. Version C was a more modest assumption.

51. Other Successful Predictions of Climate Models
More warming at night than day
Most warming in Arctic than anywhere else (especially during winter)
Least warming in/around Antarctica
Wet regions get wetter, subtropical ocean regions dry
Tropopause (at the top of the weather layer of the atmosphere) moves upward
Large scale tropical circulations weaken

52. Summary: Climate Models
Are complicated codes written by large teams of scientists. There are several dozen different models. Comparing them offers another means of verification.
Are composed of equations that describe fluid motions and have parameterizations of small scale processes involving clouds, glacial calving, plants processing moisture, etc
Are strenuously tested and have been shown to give reliable forecasts.
Differ from weather models because the initial conditions are mostly unimportant. Instead energy balance is critical. They produce storms but they are not in sync with reality. Only their statistics are relevant. 52