Transcript Examples of decadal climate prediction

Decadal Climate Prediction
Jochem Marotzke
Max Planck Institute for Meteorology (MPI-M)
Centre for Marine and Atmospheric Sciences
Hamburg, Germany
Outline




A curious apparent paradox…
Seamless prediction of weather and climate
Examples of decadal climate prediction
Ocean observations and decadal prediction
A curious apparent paradox…



We confidently predict weather one week into
the future…
We confidently state that by 2100,
anthropogenic global warming will be easily
recognisable against natural climate
variability…(cf., IPCC simulations)
Yet we make no statements about the climate
of the year 2015
Two types of predictions


Edward N. Lorenz
(1917–2008)
Predictions of the 1st kind




Initial-value problem
Weather forecasting
Lorenz: Weather forecasting
fundamentally limited to
about 2 weeks
Predictions of the 2nd kind




Boundary-value problem
IPCC climate projections
(century-timescale)
No statements about
individual weather events
Initial values considered
unimportant; not defined
from observed climate state
Can we merge the two types of prediction?

John von Neumann wrote in 1955: “The
approach is to try first short-range forecasts,
then long-range forecasts of those properties of
the circulation that can perpetuate themselves
over arbitrarily long periods of time....and only
finally to attempt forecasts for medium-long
time periods.”
Seamless prediction of weather and climate

“It is now possible for WCRP to address the
seamless prediction of the climate system from
weekly weather to seasonal, interannual,
decadal and centennial climate variations and
anthropogenic climate change.” (WCRP 2005)
Seamless prediction of weather and climate




Combination of predictions of first and second
kind – start from observed climate state; include
change in concentrations of greenhouse gases
and aerosols
Already practiced in seasonal climate prediction
(El Niño forecasts)
In decadal prediction, anthropogenic climate
change and natural variability expected to be
equally important
Atmosphere loses its “memory” after two weeks
– any predictability beyond two weeks residing
in initial values must arise from slow
components of climate system – ocean,
cryosphere, soil moisture…
Seamless prediction of weather and climate


Data assimilation & initialisation techniques
(developed in weather & seasonal climate
prediction) must be applied to ocean,
cryosphere, soil moisture
Also “imported” from seasonal climate
prediction: building of confidence (“validation”)
of prediction system, by hindcast experiments
(retroactive predictions using only the
information that would have been available at
the time the prediction would have been made)
Policy relevance of decadal climate prediction



“Long-term” planning in industry, business &
public sector overwhelmingly occurs on the
decadal timescale
Adaptation planning to climate change
overwhelmingly occurs on the decadal
timescale
Clear that, in addition to the multi-decadal
mitigation planning & very-long term
perspective, decadal timescale is crucial
Examples of decadal climate prediction

1.
2.
3.
Differences arise from models used, but mainly
(?) from the method by which the ocean
component of coupled model is initialised:
“Optimal interpolation” (Hadley Centre,
European Centre for Medium-Range Weather
Forecasts)
Forcing of sea surface temperature (SST) in
coupled model toward observations (IFMGEOMAR & MPI-M)
Using 4-dimensional ocean synthesis (ECCO)
to initialise ocean component (MPI-M & UniHH)
Hadley Cntr. prediction, global-mean surface temp.
D. M. Smith et al., Science 10 August 2007
IFM-GEOMAR & MPI-M decadal prediction
Keenlyside
et al. (Nature
2008)
Decadal-mean global-mean surface temp.
Keenlyside et al. (Nature 2008)
MPI-M & UniHH prediction: N-Atl. SST
Annual
Pentadal
Assimilation
HadISST
Hind- & Forecasts
Free model
Decadal
Pohlmann et al. (2008)
MPI-M & UniHH prediction: Global SST
Annual
Assimilation
Pentadal
HadISST
Hind- & Forecasts
Free model
Decadal
Pohlmann et al. (2008)
MPI-M & UniHH prediction: N-Atl. SST
HadISST
Forecasts
Free model
Pohlmann et al. (2008)
Organisation of decadal prediction (WCRP)

Decadal prediction is a vibrant effort if one
considers the focus on


Ocean initialisation
Atlantic
We need to develop broader scope concerning
 Areas other than the Atlantic
 Roles in initialization of:




Cryosphere
Soil moisture
Stratosphere
The science of coupled data assimilation &
initialisation has not been developed yet
Ocean observations and decadal prediction




Initialisation of ocean component of coupled
models is the most advanced initialisation
aspect of decadal prediction
Yet, methodological uncertainties are huge
Example: Meridional Overturning Circulation
(MOC) in the Atlantic
Take-home message: Comprehensive and
long-term in-situ and remotely-sensed
observations are crucial
North Atlantic Meridional Overturning Circulation
(a.k.a. Thermohaline Circulation)
Quadfasel (2005)
MOC at 25N in ocean syntheses (GSOP)
Bryden et al. (2005)
ECMWF
Monitoring the Atlantic MOC at 26.5°N
(Marotzke, Cunningham, Bryden, Kanzow, Hirschi, Johns, Baringer, Meinen, Beal)
Data recovery :
April, May, Oct. 2005;
March, Mai, Oct., Dec.
2006, March, Oct 2007,
March 2008
Church (SCIENCE,
17. August 2007)
Monitoring the Atlantic MOC at 26.5°N
(Marotzke, Cunningham, Bryden, Kanzow, Hirschi, Johns, Baringer, Meinen, Beal)
Monitoring the Atlantic MOC at 26.5°N
(Marotzke, Cunningham, Bryden, Kanzow, Hirschi, Johns, Baringer, Meinen, Beal)
First observed MOC time series, 26.5N Atlantic
Florida
Current
MOC
Ekman
Geostrophic
upper
midocean
S. A. Cunningham et al., Science (17 August 2007)
Modelled vs. observed MOC variability at 26.5N
Correlation
RMS variability
Observations
ECCO (Ocean Synthesis)
ECHAM5/MPI-OM
Baehr et al. (2008)
Update – 2.5 years of MOC time series at 26.5 N
Kanzow et al. (2008, in preparation)
Outlook – MOC monitoring at 26.5N




Dec. 2007: NERC will continue the funding
for MOC monitoring until 2014
Transformation into operational array must
take place during that period
Data need to enter data assimilation system,
to be used in initialising global coupled
climate models
Symbiosis of sustained observations and
climate prediction (analogy to atmospheric
observations and weather prediction)
Conclusions and outlook (1)



Climate prediction up to a decade in advance is
possible, as shown by predictive skill of early,
relatively crude efforts
Desirable: multi-year seasonal averages, several
years in advance, on regional scale
Sustained (operational-style) observations crucial
Conclusions and outlook (2)
Large potential for methodological improvement:
 Initialisation beyond ocean-atmosphere
(cryosphere, soil moisture)
 Development of coupled data assimilation
(challenge: disparate timescales)
 Provision of uncertainty estimate by ensemble
prediction – challenge:




Construct ensemble spanning range of uncertainty in
initial values;
Poorly known which processes dominate error growth
on decadal timescale)
Increase in model resolution for regional aspects
Vast increase in computer power required
Thank you for your
attention