Download Climate models development Climate models development

Climate models development
Dr. A. Kattenberg, KNMI, De Bilt
• What do we need climate models for?
• Climate modeling is Earth System modeling
• How do these models work?
• How good are they?
• Climate models development?
TuE Energydays 2011
1
Radiation balance of
the climate system
Incoming solar radiation
342 Wm-2
reflected solar radiation
102 Wm-2 (about 30%)
emitted infrared radiation
240 Wm-2
(corresponding to -18°C)
Earth surface is 15°C, this is 33°C ‘too warm’!
TuE Energydays 2011
2
greenhouse effect qualitatively
•
•
•
•
•
Sunlight heats the Earth
Earth re-radiates this energy in IR
Greenhouse gases absorb IR: energy
cannot escape as radiation
No greenhouse effect: -18 °C
with greenhouse eff.: +15 °C
enhanced greenhouse effect: extra
raise of surface temperature
no
natural enhanced
greenhouse greenhouse greenhouse
effect
effect
effect
TuE Energydays 2011
3
Greenhouse effect quantitatively
AR4, FAQ 1.1, Fig. 1
TuE Energydays 2011
Greenhouse gas concentrations
are on the increase
TuE Energydays 2011
5
Global temperature:
recent past & future
TuE Energydays 2011
6
Earth’s climate system
7
Radiation balance at the top of the atmosphere
TuE Energydays 2011
Really “the Earth System”
TuE Energydays 2011
8
Components in
Earthsysteem models:
Atmosphere (physics)
Oceans (physics)
Sea ice (physics)
Landsurface (physics)
New:
Atmospheric chemistry
Biology en biogeochemistry
(plants, nutrients)
Future:
Socio-economic factors
TuE Energydays 2011
9
Structure of a general circulation model
In each gridbox:
Newton (F=m x a)
+
Thermodynamics,
physics, …
… millions of lines of
10
softwarecode
TuE Energydays 2011
Atmospheric Processes
Atmospheric
prognostic
variables
Atmospheric
processes
Surface
exchanges
Surface
variables
Wind
Temperature
Mixing
Friction
Surface
roughness
Humidity
Cloud Water/Ice
Condensation/
Evaporation
Radiation
Sensible
heat flux
Surface
temperature
Precipitation
Latent
heat flux
Snow
Soil moisture
Melting
11
TuE Energydays 2011
Supercomputer power and data
are limiting factors
12
TuE Energydays 2011
Modeling dilemma …
Resolution
Complexity
13
TuE Energydays 2011
Relevant temporal and spatial scales
minute
day
year
century
Global climate
10000
CO2 variations
Horizontal length scale (km)
monsoon
El Niño
synoptic-scale
system
1000
Ocean
circulation
soil moisture
100
sea breeze
soil erosion
thunderstorms
10
interseasonal
vegetation
cumulus
1
thermals
turbulence
100
102
104
106
108
Time scale (seconds)
1010
14
TuE Energydays 2011
Relevant subsystems: model complexity
Atmosphere
Models
The Earth System
Unifying the Models
Climate / Weather
Models
The Predictive
Earth System
Hydrolog
y
Process
Models
Ocean
Models
Land
Surface
Models
Natural Hazard
Prediction
Terrestrial
Biosphere
Models
Megaflops
Gigaflops
Teraflops
2000
Petaflops
2015
16
TuE Energydays 2011
The “unnoticed”” revolution
Ensemble
forecasts
suggest
similar skills
The “unnoticed”” revolution
1.5 km resolution
simulation of Typhoon
Megi
Courtesy of UK
MetOffice
The “unnoticed”” revolution
Climate model skill
Better skill
Reichler and Kim, 2008
CMIP5 simulations
MISU/SMHI (Brodeau,Wyser)
Global mean temperature
ME industrial simulations
287.5
MetEireann (T. Semmler)
287
Temperature [K]
286.5
m e01
m e01 10-year r unni ng
m e02
m e02 10-year r unni ng
m e03
m e03 10-year r unni ng
m e04
m e04 10-year r unni ng
m e05
m e05 10-year r unni ng
ensem bl e
ensem bl e 10-year r unni ng
286
285.5
285
284.5
1854 1862 1870 1878 1886 1894 1902 1910 1918 1926 1934 1942 1950 1958 1966 1974 1982 1990 1998
1850 1858 1866 1874 1882 1890 1898 1906 1914 1922 1930 1938 1946 1954 1962 1970 1978 1986 1994 2002
Year
IM/UL (Dutra?)
DMI (Yang)
NB. Thanks to Klaus Wyser
for excellent coordination!
The “unnoticed”” revolution
IPCC, 2007
Past climate forcings
23
TuE Energydays 2011
EC-Earth consortium
The Netherlands
KNMI, U Utrecht, WUR, VU. SARA
Denmark
Ireland
DMI, Univ
Copenh
MetEireann,
UCD, ICHEC
Portugal
Switzerland
IM, U Lisbon
ETHZ, C2SM
Spain
Sweden
SMHI, Lund U, Stockholm
U, IRV
Germany
IFM/GEOMAR
AEMET, BSC, IC3
Norway
MetNo, NTNU
Belgium
UCL
Italy
ICTP,CNR,
ENEA?
Steering group: W. Hazeleger (KNMI), C. Jones (SMHI), J. Hesselbjerg, Christensen (DMI), R. McGrath
(Met Eireann), observer E. Kallen (ECMWF), NEMO-representative
ECMWF
EC-Earth
EC-Earth
Vx
IFS Cycle 36
NEMO-3
Univ/institutes
Atmospheri
c chemistry
Dynamic
Vegetation
EC-Earth
V2
~3 yr
Snow
Land use
Aerosols
EC-Earth
V1
IFS Cycle 31
NEMO-2
What to expect from decadal
predictions?
•
Prognostic potential predictability: ensemble spread in relation to total
variance
PPP=1−
N
M
1
2
[
X
(
t
)
X
(
t
)]
−
∑ ∑ ij
j
N(M −1) j=1 i=1
σ2
With Xij is the ith member of jth ensemble, N is number ensembles, M number
of ensemble members
Griffies and Bryan 1997, Collins et al, Pohlmann et al
Prognostic potential predictability in ECEarth (T2m, yr 1-10; without trend)
T. Koenigk, SMHI, pers. comm.
Multi-model decadal predictions
2 meter temperature multi-model anomaly correlation
2-5 year lead time averaged, without trend.
Using EU-Ensembles data.
Model complexity
The Seventies and Eighties
Land
Atmosphere
Courtesy Christian Jacob & Pier Siebesma
Model complexity
The Eighties and Nineties
Aerosols
Ocean and Sea Ice
Land
Atmosphere
Model complexity
Today
Services
Chemistry
Carbon
Aerosols
Ocean and Sea Ice
Land
Atmosphere
Solution 1
Strengthen the
foundations of
the existing
house.
?
?
Physical Model
developer
Tasmanian Devil
“An endangered species is a population of organisms which is at risk of becoming
extinct because it is either few in numbers, or threatened by changing environmental
or predation parameters.” - Wikipedia
?
Physical Model
developer
Tasmanian Devil
Endangered Species
An endangered species program
Improve the habitat
Breed the species
Modelling Centres
Academia
• Open the models to the
•
•
community
More internal collaboration
Reward system
• Stronger formal links
• Joint PhD to guaranteed
• Special scholarships
Postdoc positions
• Model development chairs
• Strategic partnerships with • Reward system
funding programmes
Model complexity
Today
Services
Chemistry
Carbon
Aerosols
Ocean and Sea Ice
Land
Atmosphere
Solution 2
Build a new,
more modern
house, with
foundations of
the required
strength.
A “Manhattan project””
• Improved models are key to achieving the skill of
predictions society is asking of us.
• This calls for a “Manhattan-style” project on
developing the best model we can today.
• The main purposes of such a project would be to
advance the science of modelling and to
demonstrate the effect on key predictions.
• Must build a new model using modern ideas (e.g.,
stochastic approaches, high-efficiency and highresolution dynamical cores, ...).
A “Manhattan project””
Improved Models for better
predictions
Improve physics
Identify and improve
key processes
Estimate
uncertainties
Stochastic physics
Reduce physics
influence
Ensembles
A Manhattan project for model development
Increased
resolution
A “Manhattan project””
• Model development should not be undertaken in
isolation from prediction
• Option 1: Link the project with a few existing
modelling centres
• Option 2: A new centre, say for seasonal to decadal
prediction
• Could be centralised or distributed
Model development is a community effort
Jakob, BAMS
2010
Application
Model
Overall
assessment
NWP; seasonal;
climate
Tuning (important but limited insight)
Design model
improvements
Great insight but of potentially
limited importance
Perform process
studies (model +
observations)
Data community
Model user/ evaluation
community
Find processes
and phenomena
of relevance
Select suitable
process studies
Model development community
Climate models development
• Weather and climate models underpin some of mankind’s
greatest endeavours. They save lifes. They save property.
They affect all aspects of society.
• Improvements in forecasts and projections have been
underpinned by improvements in models - Future
improvements require renewed and increased investment in
basic model development.
• Models have become increasingly complex, but some key
issues have not been resolved. We need both an
endangered species program for model developers and our
own “Manhattan”-style project to successfully implement the
seamless prediction paradigm.
• The time is right!