Frank Selten, Global Climate Dept, KNMI - Universiteit...

64
Frank Selten, Global Climate Dept, KNMI Climate modelling

Transcript of Frank Selten, Global Climate Dept, KNMI - Universiteit...

Page 1: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Frank Selten, Global Climate Dept, KNMI

Climate modelling

Page 2: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Frank Selten, Global Climate Dept, KNMI

Climate modelling

Page 3: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

What is a climate model ?

A climate model is a quantitative mathematical description of aspects of the earth’s climate system

Page 4: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Purpose of a climate model

A climate model is used to simulate the behavior of the earth’s climate system in order to

increase our understanding of how the climate system workssimulate past climate fluctuations to help interpret historical and paleo observationssimulate the future climate based on scenario’s for emissions of green house gasses

Page 5: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Present state

Specify external forcings according to the date

Model calculates rate of change (tendency)

State one timestep (30 minutes) later

Principle of a climate simulation

Page 6: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Earth system: our natural environment

ecosphere

anthroposphere

socio-economics

values

beliefs

biosphere geosphere

attitudes

atmospherehydrospherecryosphereastenosphere

vegetationsea biology

Page 7: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Earth system: our natural environment

ecosphere

anthroposphere

socio-economics

values

beliefs

biosphere geosphere

attitudes

atmospherehydrospherecryosphereastenosphere

vegetationsea biology

Energy Balance Model

Page 8: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Energy balance model

Total incoming radiation = Total outgoing radiation

Solar radiationThermal radiation

σT4=annual and global mean absorbed solar radiation ε

Page 9: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Energy balance model

σT4=annual and global mean absorbed solar radiation ε

πR S (1-α) = 4πR εσT2 40

R :S :α :ε :σ :

0

2

radius of the earth (6378 km)solar constant (1370 W/m )2

average albedo (0.32)emissivity (≈ 1)Stefan-Boltzmann constant (5.67E-8 W/m /K )2 4

T = -20 degree Celsius

Page 10: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Without / with Greenhouse Effect: -20 / 14 degrees Celsius

EBM: the Greenhouse effect ...

CO2H20

more

CO2

Page 11: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Earth is warming ...

Page 12: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

time →1902

year

summer

autumn

spring

winter

Colder than averageWarmer than average

2002

Also in De Bilt ...

Page 13: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

And Carbon Dioxide is to blame ...

Page 14: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

The global energy balance ...

Page 15: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Δ temperatureΔ radiation

Current ‘best’ estimate:

around 0.8 K/W/m2

or

3K for CO2 doubling

Role of feedbacks

Climate sensitivity:

Page 16: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

The warming is not uniform ...

Page 17: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

The physical climate system ...

Page 18: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Global climate models solve the (thermo-)dynamical equations on a computational grid

Current resolutions: 100 km and 40 layers in the vertical

Global climate models solve the (thermo-)dynamical equations on a computational grid

Current resolutions: 100 km and 60 layers in the vertical

Page 19: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Primitive (hydrostatic) equations

forMomentum equations

Sub-grid model :“physics”

Numerical diffusion

Page 20: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

hydrostatic equations

Thermodynamic equation

Moisture equation

Note: virtual temperature Tv instead of T from the equation of state.

Page 21: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

hydrostatic equations

Continuity equation

Vertical integration of the continuity equation in hybrid coordinates

Page 22: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Resolution problem• So far : We derived a set of evolution equations

based on 3 basic conservation principles valid at the scale of the continuum : continuity equation, momentum equation and thermodynamic equation.• What do we want to (re-)solve in models based on

these equations?

grid

scal

e

reso

lved

scal

e (?

)

The scale of the grid is much bigger than the

scale of the continuum

Text

Page 23: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

23

~10−6 m - 1m

~107 m ~105 m

~103 m

The planetary scaleCloud cluster scale

Cloud scaleCloud microphysical scale

The climate system : A truly multiscale problem

Page 24: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

10 m 100 m 1 km 10 km 100 km 1000 km 10000 km

turbulence Cumulusclouds

Cumulonimbusclouds

Mesoscale Convective systems

Extratropical Cyclones

Planetary waves

Large Eddy Simulation (LES) Model

Cloud System Resolving Model (CSRM)

Numerical Weather Prediction (NWP) Model

Global Climate ModelSubgrid

No single model can encompass all relevant processes

DNS

mm

Cloud microphysics

Page 25: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

25

Grid-box size is limited by computational capability

Processes that act on scales smaller than our grid box will be excluded from the solutions.

We need to include them by means of parametrization (a largely statistical description of what goes on “inside” the box).

Similar idea to molecules being summarized statistically by temperature and pressure, but much more complex!

Parametrization

Page 26: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

What is a climate model ?

Parameterization of unresolved processes biggest source of model errorsDifferent climate models differ in these descriptionsHow to optimally combine the predictions of the different models?

Page 27: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

A multi-model ensemble method that combines imperfect models through

learningLeonie van de Berge, Mathematical Institute UtrechtFrank Selten*, Royal Netherlands Meteorological InstituteWim Wiegerinck, Radboud University NijmegenGreg Duane, University of Colorado

Earth System Dynamics, 2011

*Global Climate dept: 12 staff 14 postdocs/phd students Climate research: 120 scientists

Page 28: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Multi model simulation of real complex systems

Reality

• climate system• ecological systems• human brain• organisms• economic systems

Model 1 Model 2 μοδελ 4

Observed data

Model 1data Model 2 data Model 3 data μοδελ 4 data

Combined data

Usually some form of a weighted average

{

An ensemble of “imperfect models”

Model 3

Page 29: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Climate models

• order 20 global coupled climate models

• have improved performance over time

• but are not perfect

• are used to simulate response to different scenarios of future emissions of greenhouse gasses

• differ in the simulation of the response

Page 30: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Coupled Model Intercomparison Project

Performance metricBased on mean squared errors in time mean global temperatures, winds, precipitation, ....

1995

2006

= index value based on multi model mean fields: outperforms individual models: why ?

Page 31: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Error in annual mean surface air temperaturesmulti model mean over all CMIP3 simulations

IPCC 2007

Page 32: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Spread in simulated climate change

IPCC 2007

Page 33: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Given present state of modeling ......

...... is this the best we can do ?

Page 34: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

I have an idea !!!

Page 35: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Multi model simulation of real complex systems

Reality

Model 1 Model 2 μοδελ 4Model 3

Observed data

Model 1data Model 2 data Model 3 data μοδελ 4 data

Combined data

Usually some form of a weighted average

An ensemble of “imperfect models”

Model 3

Page 36: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Exchange information between models while integrating

Reality

Model 1 Model 2 μοδελ 4

Observed data

Model 1data Model 2 data Model 3 data μοδελ 4 data

Combined data

multi model averaging

An interacting ensemble of “imperfect models”

Model 3

Page 37: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

How ?

• Example using the chaotic Lorenz 1963 model

Page 38: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Lorenz 1963 model

has for standard parameter values:

a chaotic solution:

Page 39: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar
Page 40: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar
Page 41: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Perfect model approach

• model with standard parameter values ⇒ truth

• perturb parameter values to create an ensemble of imperfect models

• exchange of information between the imperfect models takes the form of linear “nudging terms”

Page 42: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Interacting ensemble of imperfect models

where k indexes the imperfect models

Effectively a new dynamical system is created, “a super model”,

with adjustable connection coefficients C

For k=3 we have 18 coefficients

Page 43: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Use data from the truth to learn the connection coefficients

Minimize Cost function:

Page 44: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Imperfect models unconnected

Model 1: fixed point Model 2: fixed point Model 3: strange attractor

Page 45: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

“a” supermodel solutionfrom two different view points

Page 46: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

The three connected models fall into an approximate synchronized motion

Model 1 Model 2 Model 3

Put differently: the models form a consensus

Synchronization of chaotic systems is a well-known phenomenon

Page 47: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

The vertical discretisation is given in Fig. 5. The vorticity equation is applied to the200 (level 1), 500 (level 2) and the 800 hPa level (level 3), the heat equation is applied tothe 650 and 350 hPa level. Combined these two equations yield an equation expressingthe conservation of quasi-geostrophic potential vorticity (PV) in the absence of forcingand dissipation. In discretised form it reads

q̇1

= �J ( 1

, q1

)�D1

( 1

, 2

) + S1

q̇2

= �J ( 2

, q2

)�D2

( 1

, 2

, 3

) + S2

q̇3

= �J ( 3

, q3

)�D3

( 2

, 3

) + S3

, (19)

where the index i = 1, 2, 3 refers to the pressure level. Here PV is defined as

q1

= r2 1

�R�2

1

( 1

� 2

) + f

q2

= r2 2

+ R�2

1

( 1

� 2

)�R�2

2

( 2

� 3

) + f

q3

= r2 3

+ R�2

2

( 2

� 3

) + f(1 +h

H0

), (20)

where R1

(=700 km) and R2

(=450 km) are Rossby radii of deformation appropriate tothe 200-500 hPa layer and the 500-800 hPa layer, respectively and H

0

is a scale height.In eqns.( 19), D

1

, D2

, D3

are linear operators representing the e↵ects of Newtonian relax-ation of temperature, linear drag on the 800 hPa wind (with drag coe�cient dependingon the nature of the underling surface), and horizontal di↵usion of vorticity and temper-ature. The temperature relaxation has a radiative time scale of 25 days, the linear dragdamps the low-level wind on a spin down time scale of 3 days over the oceans, about 2days over low-altitude land and about 1.5 days over mountains above 2 km; a (stronglyscale-selective) horizontal di↵usion damps harmonics of total wavenumber 21 on a 2-daytime scale.The PV source terms S

i

are calculated from observations as the opposite of the averagePV tendencies obtained by inserting observed daily winter time stream function fieldsinto a version of eqns. (19) in which these terms are omitted.A Galerkin projection of eqns. (19) onto a basis of spherical harmonics truncated attotal wavenumber 21 leads to a system of 1449 coupled ordinary di↵erential equationsfor the 483 coe�cients of the spherical harmonical functions at the three levels. Simi-lar as for the T21 barotropic model (see section 1.5) the spectral transform method isimplemented to evaluate the quadratic interaction terms in order to save computationtime. In this case the restriction to modes with m + n odd is not made and the modelsimulates the flow at the Southern hemisphere as well.

The mean state and the variability are surprisingly realistic as compared to the ob-served wintertime flow in the Northern hemisphere Corti et al (1997). The simulation isless realistic in the summer hemisphere. In addition, the model displays regime behaviornot unlike the observations (Selten and Branstator, 2004). The issues for super modelingin the context of this model concern the density and form of the connections and theability of the super model to reproduce the regime behavior.

17

A global atmosphere model

where Pm,n

(µ) denote associated Legendre polynomials of the first kind. The Legendrepolynomials are normalized in the same way as in Machenhauer (1991). The sphericalharmonics are eigenfunctions of the Laplace operator:

�Ym,n

(�, µ) = �n(n + 1)Ym,n

(�, µ) (12)

The expansion is triangularly truncated at wavenumber 21:

(�, µ, t) =21X

n=1

+nX

m=�n

m+n=odd

m,n

(t)Ym,n

(�, µ) (13)

The restriction to modes with m + n odd excludes currents across the equator. Thismakes the model hemispheric. After the Galerkin projection, the discretized model con-sists of 231 coupled ordinary di↵erential equations for the coe�cients of the sphericalharmonics. The orography is also projected onto the spherical harmonics and truncatedat T21. In principle the set of ODE’s has the same structure as the 6D barotropic modelin the previous section (eqns. 8) with constant, linear and quadratic terms. However,the number of quadratic interaction terms makes tendency evaluations of this T21 modelcomputationally expensive. A cheaper solution is implemented to calculate the quadraticterms on a computational grid in physical latitude-longitude space using fast transfor-mation routines to transform the spectral representation to the grid representation andvice-versa (see Machenhauer (1991)).We used a dataset of 10 winters of ECMWF daily analysis of 300 hPa vorticity fields

to calculate the forcing. The fields were multiplied by a factor of 0.6 to approximate the500 hPa level. The dataset covers the winter months December, January and Februaryfrom 1981 to 1991. The scale height is set to 10 km and A

0

is given a value of 0.2.The Ekman damping time scale is set to 15 days and the strength of the scale selectivedamping is such that wavenumber 21 is damped at a time scale of 3 days. The forcingis calculated from the ECMWF dataset according to:

⇣⇤ = J ( cl

, ⇣cl

+ f + h) + k1

⇣cl

� k2

�3⇣cl

+ J ( 0, ⇣ 0) (14)

where cl

, ⇣cl

is the observed climatological winter mean state computed from the 10winters and 0, ⇣ 0 are deviations of the 10 days running mean from the climatologicalmean. The model climatic mean will resemble the observed climatological winter meanstate if it is capable of producing a mean transient eddy forcing similar to this observedlow frequency transient eddy forcing. With this forcing field, we integrated the modelfor 2000 days and analysed its variability. It turned out that the model variability wasrather weak compared to the observed low frequency variability. From this model runwe determined the model transient eddy forcing, with respect to the observed clima-tological mean state, and replaced in ⇣⇤ the observed transient eddy forcing with themodel transient eddy forcing. We integrated the model a second time for 2000 daysand calculated its transient eddy forcing again. We integrated the model a third time

11

Solved by the spectral method using spherical harmonics as basis functions

The vertical discretisation is given in Fig. 5. The vorticity equation is applied to the200 (level 1), 500 (level 2) and the 800 hPa level (level 3), the heat equation is applied tothe 650 and 350 hPa level. Combined these two equations yield an equation expressingthe conservation of quasi-geostrophic potential vorticity (PV) in the absence of forcingand dissipation. In discretised form it reads

q̇1

= �J ( 1

, q1

)�D1

( 1

, 2

) + S1

q̇2

= �J ( 2

, q2

)�D2

( 1

, 2

, 3

) + S2

q̇3

= �J ( 3

, q3

)�D3

( 2

, 3

) + S3

, (19)

where the index i = 1, 2, 3 refers to the pressure level. Here PV is defined as

q1

= r2 1

�R�2

1

( 1

� 2

) + f

q2

= r2 2

+ R�2

1

( 1

� 2

)�R�2

2

( 2

� 3

) + f

q3

= r2 3

+ R�2

2

( 2

� 3

) + f(1 +h

H0

), (20)

where R1

(=700 km) and R2

(=450 km) are Rossby radii of deformation appropriate tothe 200-500 hPa layer and the 500-800 hPa layer, respectively and H

0

is a scale height.In eqns.( 19), D

1

, D2

, D3

are linear operators representing the e↵ects of Newtonian relax-ation of temperature, linear drag on the 800 hPa wind (with drag coe�cient dependingon the nature of the underling surface), and horizontal di↵usion of vorticity and temper-ature. The temperature relaxation has a radiative time scale of 25 days, the linear dragdamps the low-level wind on a spin down time scale of 3 days over the oceans, about 2days over low-altitude land and about 1.5 days over mountains above 2 km; a (stronglyscale-selective) horizontal di↵usion damps harmonics of total wavenumber 21 on a 2-daytime scale.The PV source terms S

i

are calculated from observations as the opposite of the averagePV tendencies obtained by inserting observed daily winter time stream function fieldsinto a version of eqns. (19) in which these terms are omitted.A Galerkin projection of eqns. (19) onto a basis of spherical harmonics truncated attotal wavenumber 21 leads to a system of 1449 coupled ordinary di↵erential equationsfor the 483 coe�cients of the spherical harmonical functions at the three levels. Simi-lar as for the T21 barotropic model (see section 1.5) the spectral transform method isimplemented to evaluate the quadratic interaction terms in order to save computationtime. In this case the restriction to modes with m + n odd is not made and the modelsimulates the flow at the Southern hemisphere as well.

The mean state and the variability are surprisingly realistic as compared to the ob-served wintertime flow in the Northern hemisphere Corti et al (1997). The simulation isless realistic in the summer hemisphere. In addition, the model displays regime behaviornot unlike the observations (Selten and Branstator, 2004). The issues for super modelingin the context of this model concern the density and form of the connections and theability of the super model to reproduce the regime behavior.

17

Page 48: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

t = 0

s, q

T,t

s, q

T,t

s, q

0 hPa

200 hPa

350 hPa

500 hPa

650 hPa

800 hPa

Ps

0

1

2

3

4

5

Figure 5: Vertical discretisation of the three-level model.

16

Page 49: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar
Page 50: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar
Page 51: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

-6

-4

-2

0

2

4

0 10 20 30 40 50

strea

mfun

ction

time [days]

model 1 NLmodel 2 NLmodel 1 SAmodel 2 SA

Synchronization between two identical baroclinic spectral T21QG models on the sphere

Streamfunctions at 500 hPaat initial time after 60 days

dq2mn

dt= .......− ci q2

mn − q1mn( )

dq1mn

dt= .......− ci q1

mn − q2mn( )

Page 52: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar
Page 53: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar
Page 54: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Super model solutions are not unique:cost function F(C) has isolated local minima

Page 55: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Cross sections of the cost function

• convergence for increasing size of training set• for some connection constants, the cost

function is flat: family of solutions

Page 56: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

But the solutions differ in quality

8

Page 57: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

But the solutions differ in quality

Page 58: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Can the super model simulate climate change?

since super model is trained on present day climate ...

Doubling the parameter ρ from 28 to 58: supermodel simulates change well

Page 59: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Imperfect models Lorenz 1984

Model 1: periodic orbit Model 2: fixed point Model 3: periodic orbit

x: strength of westerlies

y,z: sine and cosine phase of a wave

Page 60: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

“a” supermodel solutionfrom two different view points

Page 61: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Questions• Are other forms of the connections more effective?

• How many connections are required ?

• Which variables to be connected and how often ?

• How much data is needed for the learning ?

• Are there more effective learning strategies ?

• How to handle the slow oceanic time scales ?

• What if reality falls outside of the model class ?

• Does the supermodel also perform well in a changing climate ?

Page 62: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

Questions• Can we identify identical state variables in the

different models ?

• Instead of connecting state variables, is connecting the physical tendencies a good idea?

• Do balances and conservation laws place restrictions on the connections ? Similar issues play a role in data-assimilation ...

• Is it computationally feasible to run an interconnected ensemble of climate models?

• Is it possible to choose connections on the basis of insight, without learning ?

• .........

Page 63: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

• Can the super models synchronize with the truth?

But the solutions differ in quality

Nudging to observations

Page 64: Frank Selten, Global Climate Dept, KNMI - Universiteit Utrechtsleij101/Opgaven/Caleidoscoop2/materiaal/... · ecosphere e socio-economics values beliefs ... Thermal radiation Solar

• Imperfect models do not synchronize•Perfect model synchronizes for n=3•Super models synchronize for n=11 and n=13•Distance between model and observations vary with time:

But the solutions differ in quality