msc module mtmw14 : numerical modelling of atmospheres and oceans
DESCRIPTION
MSc Module MTMW14 : Numerical modelling of atmospheres and oceans. Week 3: Modelling the Real World 3.1 Staggered time schemes (semi-implicit) 3.2 Physical parameterisations 3.3 Ocean modelling. Course content. Week 1: The Basics 1.1 Introduction - PowerPoint PPT PresentationTRANSCRIPT
(c) 2004 [email protected] 1
MSc Module MTMW14: Numerical modelling of atmospheres and oceans
Week 3: Modelling the Real World
3.1 Staggered time schemes (semi-implicit)3.2 Physical parameterisations3.3 Ocean modelling
2
Course content
Week 1: The Basics1.1 Introduction1.2 Brief history of numerical weather forecasting 1.3 Dynamical equations for the unforced fluid
Week 3: Modelling the real world3.1 Physical parameterisations3.2 Ocean modelling3.3 Staggered time schemes and the semi-implicit method
Week 4: More advanced spatial methods4.1 Staggered horizontal and vertical grid discretisations4.2 Lagrangian and semi-lagrangian schemes4.3 Series expansion methods
Week 5: Final thoughts5.1 Revision5.2 Test5.3 Survey of state-of-the-art coupled climate models
3
3.2 Physical parameterisations
],;[][ tFQt
Q is the fluid dynamics term (dynamical core)
F is forcing and dissipation due to physical processes such as:
• radiation• clouds• convection• horizontal and vertical mixing/transport• soil moisture and land surface processes• etc.
The atmosphere and oceans are FORCED DISSIPATIVE fluid dynamical systems:
4
“Earth System” Science, NASA 1986
From Earth System Science – Overview, NASA, 1986
5
In the Hollow of a Wave off the Coast at Kanagawa", Hokusai
3.1 Unresolved process: horizontal mixing
6
3.1 Small-scale mixing
Two approaches:
• Direct Numerical Simulation (DNS)try to resolve all spatial scales using a very high resolution model.
• Large-Eddy Simulation (LES)model the effects of sub-grid scale eddies as functions of the large-scale flow (closure parameterisation).
][]'[
'
FQ
''
0'
'
babaab
a
aa
aaa
:averaging sReynold' on Tips
7
3.1 Horizontal mixing in vorticity eqn
)(''
0)''.().(
0))')('.(('
0).(
fu
uut
uutt
ut
:by eddies grid-sub to due fluxvorticity mean the separameteri totry then
:boxes grid over average
Or more correctly model fluxes due to unresolved scales as f(.) + random noise rather than just f(.). (STOCHASTIC PHYSICS)
8
1.3 Some mixing parameterisations
DiffusionAssume that vorticity flux due to eddies acts to reduce the large-scale vorticity gradient:
Note: Mixing isn’t always down-gradient. For example, vorticity fluxes due to mid-latitude storms act to strengthen the westerly jets!
)viscosity! molecular (NOT(d) viscosity"eddy "
2)''.(
''
u
u
9
1.3 Some mixing parameterisations
Hyper-diffusion
Diffusion is found to damp synoptic features too much. More scale-selective damping can be obtained using:
Note: Widely used in spectral models.
klmm
m
lk
etcm
u
~)()(
.,4,3,2
)()''.(
222
2
: space spectral in form Simple
10
1.3 Some mixing parameterisations
Strain-dependent viscosity (Smagorinsky 1963)Put more diffusion in regions that have more large-scale strain:
Note: Diffuses more strongly in high strain zones (e.g. between cyclones) but is computationally expensive to implement.
12
2
2
][
2
1)2.0(
TL
Dimensions
y
v
x
u
11
3.1 Spectral blocking and instabilityIn 2-d turbulence and the atmosphere and oceans, enstrophy cascades to ever smaller scales:
This cascade can’t continue for ever in numerical models and so in the absence of any mixing parameterisations, energy builds up on the smallest scales. This is known as spectral blocking and can lead to non-linear instability:
Phillips, N. A., 1959: An example of nonlinear computational instability. In: B. Bolin (Editor), The Atmosphere and the Sea in Motion. Rockefeller Inst. Press in association with Oxford Univ.Press, New York, 501-504.
2
12
3.1 Convection schemes
Why parameterise cumulus convection?
Precipitation is caused by rising of air due to: Local convective instability Large-scale ascent (slantwise convection) Orographic influences
13
3.1 Convection schemes
Convection occurs on small spatial scales
(less than 10km) not explicitly resolvable
by weather and climate models.
It is important for
producing precipitation and releasing latent heat (diabatic heating of atmosphere).
producing clouds that affect radiation (e.g. cumulonimbus
Cb, cumulus Cu, and stratocumulus Sc)
vertical mixing of heat and moisture (and horizontal momentum)
Further reading:
J. Kiehl’s Chapter 10 in Climate System Modelling (Editor: K. Trenberth)
Emanuel, K.A. 1994: Atmospheric convection, Oxford University Press.
14
15
3.1 Summary of main points
Modelling physical processes is an important aspect of weather prediction and climate simulation. Not as simple as modelling the fluid dynamics parts.
Physical schemes take roughly the same amount as computer time as the fluid dynamics schemes.
Forcing (e.g. radiation) and dissipative processes (e.g. mixing) need to be correctly modelled.
Unresolved processes that feedback onto the large-scale flow need to be parameterised in terms of the the large-scale flow.
16
17
18
19
20
21
3.1 Summary of main points
22
23
1
Basic idea:
Represent time derivative by centered time finite difference (CT) and then put different terms of the equation on different time levels. (mixed scheme)
Why? Different equations have different stability
properties (e.g. diffusion equation is unstable for CT and advection equation is unstable for FT scheme).
Treat some of the terms implicitly in order to slow down simulated waves and get round CFL criterion (semi-implicit)
24
3.1 Staggered time schemes (semi-implicit)
Example:
2.
ut
)1(2)()()1()1(
.2
nnn
nn
uh
• stable leapfrog (centred) for advection• stable forward for diffusion• Note: both are conditionally stable
25
3.1 Semi-implicit gravity wave schemes
Key concept: slow down fast gravity waves by treating them implicitly while treating other terms explicitly
Example: shallow water equations
T
xx
x
x
vu
f
f
Q
Qt
),,(
0
0
0
26
3.1 Semi-implicit gravity wave schemes
Split the operator:
Centered scheme (explicit)
Backward scheme (fully implicit)
Mixed scheme (semi-implicit)
)()(
0
00
00
000
00
00
ravityGossbyR
f
f
Q
xx
x
x
)()1()1( )(2 nnn GRh
)1()1()1( )(2 nnn GRh
)(2 )()()1()1( nnnn GRh
27
3.1 Semi-implicit gravity wave schemes
The sparse operator 1+hG is a lot easier to invert than is 1+2hQ – obtain a simple Helmholtz equation for height field. Rank of matrix to be inverted is 1/3 that of fully implicit scheme.
Fast sparse matrix techniques can be used to invert (1+hG)
Causes gravity waves to be treated implicitly and so larger time steps can be used without running into CFL problems
ECMWF example:Ritchie …
28
3.1 Summary of main points
29
30
Winnighoff 1968, Arakawa and Lamb 1977.)(
2
1)(
)(2
1)(
)(
,2/1,2/1
,2/1,2/1
,2/1,2/1
jiy
jiy
ij
xy
jijiij
x
jijiijx
31
32
ECMWF algorithm
33
34
35
3.1 Summary of main points
36
If the lapse rate exceeds dry adiabatic, a convective adjustment restores the lapse rate to dry adiabatic, with conservation of dry static energy in the vertical. A moist convective adjustment scheme after Manabe et al. (1965) also operates when the lapse rate exceeds moist adiabatic and the air is supersaturated. (For supersaturated stable layers, nonconvective large-scale condensation takes place--see Precipitation). In moist convective layers it is assumed that the intensity of convection is strong enough to eliminate the vertical gradient of potential temperature instantaneously, while conserving total moist static energy. It is further assumed that the relative humidity in the layer is maintained at 100 percent, owing to the vertical mixing of moisture, condensation, and evaporation from water droplets. Shallow convection is not explicitly simulated.
Manabe, S., J. Smagorinsky, and R. F. Strickler, 1965: Simulated climatology of a general circulation model with a hydrologic cycle. Mon. Wea. Rev., 93, 769-798.
Arakawa, A. and W.H. Schubert, 1974: Interaction of cumulus cloud ensemble with the largescale environment, Part I. J. Atmos. Sci., 31, 674-701.
Betts, A.K. and M.J. Miller, 1986: A new convective adjustment scheme. Part I: Observational & theoretical basis. QJRMS, 112, 677-691.
Kuo, H.L., 1974: Further studies of the parameterization of the influence of cumulus convection on large scale flow. J. Atmos. Sci., 31, 1232-1240.
Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117, 1799-1800.
37