Kein Folientitel

1
Do Simplified Models Have any Useful Skill ?
Thomas Stocker
Climate and Environmental Physics University of
Bern, Switzerland www.climate.unibe.ch
Collaborators R. Knutti, F. Joos, G.-K.
Plattner, A. Schmittner
2
Content of Presentation
1. A model hierarchy 2. Simplified models
progress or confusion? 3. When are simplified
models useful?

• extensive parameter studies
• developing novel model strategies
• coupling biogeochemistry and other tracers

4. Prospects of simplified modeling
3
Hierarchy of Climate Models
4
approx. CPU time for 250-year simulation
5
Example 1
Multiple equilibrium states of the Atlantic
overturning circulation
6
One wonders whether other, quite different
states of flow are permissible in the ocean or
some estuaries and if such a system might jump
into one of these with a sufficient perturbation.
If so, the system is inherently fraught with
possibilities for speculation about climatic
change. Stommel (1961)
7
Bryan (1986)
8
Multiple Equilibria of the Atlantic MOC

• Stocker Marchal (2000)

9
Simplified models propose a unified explanation
of abrupt climate change
10
Example 2
Ocean circulation during the last glacial maximum
(20,000 years ago)
11
Atlantic MOC at the Last Glacial Maximum

• zonally averaged ocean model
• statistical dynamical atmosphere
• modified Arctic freshwater fluxes

Ganopolski et al. (1998)
12
Atlantic MOC at the Last Glacial Maximum

• 3d ocean model
• energy-moisture balance atmosphere
• modified continental run-off

Weaver et al. (1998)
13
Atlantic MOC at the Last Glacial Maximum
DWF shifted south MOC at similar strength deep
ocean less ventilated
DWF as today MOC reduced by gt 60 deep ocean
less ventilated
14

• Specific realisations of simplified models are
  too strongly influenced by model setup and
  parameter choices.
• No information on the robustness of the results.

15
What constitutes a useful contribution of
simplified models to science progress ?

• extensive parameter studies
• developing novel model strategies
• coupling biogeochemistry

16
Simplified models are useful ...
to perform extensive parameter studies and to
explore uncertainties
17
Change of the Atlantic meridional overturning

• IPCC TAR (2001, Chapter 9)

18
Greenhouse Gas Forcing
Atlantic MOC Beyond 2100
(Stocker Schmittner, 1997)
19
Threshold of North Atlantic MOC
Atlantic MOC Beyond 2100
(Stocker Schmittner, 1997)
20
Parameter dependence of MOC threshold
(Stocker Schmittner, 1997)
21
Parameter dependence of MOC threshold
(Stocker Schmittner, 1997)
22
Predictability of Atlantic MOC
Knutti Stocker (2002)
23
Simplified models are useful ...
to develop novel modeling strategies to outline
possible application by more complex models
24
IPCC WG1, SPM
25
Estimating a PDF requires 5,000 model simulations
26
Influence of mixing schemes on climate sensitivity
27
Climate sensitivity in different comprehensive
models
IPCC TAR WG1, Table 9.A1
It seems likely that models with higher
sensitivity, those predicting the most drastic
anthropogenic climate changes in the future, may
have difficulty satisfying the ocean
constraint. (Barnett et al., 2001)
28
global ocean heat uptake from the mid 50s to the
mid 90s 2?1023 J
Levitus et al. (2000)
29
Coupled model simulation including anthropogenic
forcing

• observed GHG
• direct aerosol effect
• no indirect aerosol effect

ocean heat uptake consistent for small climate
sensitivity.
Barnett et al. (2001)
30
Simulated ocean heat uptake 1955 - 1995
complete forcing
no aerosol cooling
Knutti et al. (2002)
31
How well do observations constrain climate
sensitivity?
complete forcing
Knutti et al. (2002)
32
How well do observations constrain climate
sensitivity?
no aerosol cooling
complete forcing
Knutti et al. (2002)
33
Simplified models are useful ...
to include biogeochemical cycles and various
tracers to investigate processes in the coupled
system
34
Climate model of reduced complexity

• zonally averaged ocean, 3 basins
• energy balance atmosphere
• thermodynamic sea ice model
• annual mean or seasonal cycle
• prognostic carbon cycle

35
Anthropogenic CO2 in the Atlantic
Plattner et al. (2002)
36
Outgassing of O2 in the Atlantic
Plattner et al. (2002)
37
Revision of land uptake estimates
Plattner et al. (2002)
38
Do simplified models have any useful skill?
39
Simplified models pros and cons

• positive
• great tools, efficient, lots of sensitivity
  studies
• dont need big computers
• parameterisations can be kept transparent

• negative
• implication of reduced dynamics not testable
• what was the model designed for?
• time and spatial scales?

40
Simplified models pros and cons

• contribution to progress
• exploration of parameter space
• estimation of uncertainty
• development of novel methodologies
• long-term simulation of coupled system

• not useful
• singular simulation syndrom
• data fitting
• over-parameterisation

41
Judicious use of simplified models

• simplified models cannot replace complex models.
  Reduced dynamics may, however, exhibit more
  clearly the dominating processes and feedback
  mechnisms at work.
• simplified models are useful if applied
  judiciously. Beware of singular simulations!
  Robustness of results can be tested much better
  than in more complex models.
• simplified models serve to test hypotheses, and
  explore their implications for other components
  of the climate system. They help target
  experiments with more expensive models.

42
Prospects of simplified models

• They shall not disappear! Simplified models
  remain valuable helpers in understanding the
  Earth system.
• Simplified models will be used in feasibility
  studies before expensive simulations are started.
• Simplified models will be used to interpret
  paleoclimatic data, and to test/determine
  transfer functions for proxy indicators.
• Simplified models will be increasingly used for
  long-term (millennial) simulations for past and
  future climate changes.