Nonhydrostatic IFS at ECMWF

1
Non-hydrostatic IFS at ECMWF
Mariano Hortal, Deborah Salmond, Agathe Untch,
and Nils Wedi
2
Topics

• NH-IFS stability
• physics-dynamics coupling
• NH-IFS climate and forecast performance
• Tracer transport physics-dynamics coupling
• idealized flow experiments

3
NH-IFS

• Problem with horizontal diffusion, which lead to
  noise and model failure in the stratosphere.

4
Problem with horizontal diffusion
Inconsistency in setup of horizontal diffusion in
NH at ECMWF diffusion applied on horizontal
divergence and the two new non-hydrostatic
variables was different. This did not show up as
a problem in CY30R1, however in CY31R1 it results
in noise in the horizontal divergence
When the same diffusion is applied on the NH
variables as on horizontal divergence this
noise does not develop.
5
Model failure when coupled to the physics

• Using large time-steps we encountered problems
  with bulls eyes in the temperature field near
  the surface in areas with steep orographic
  gradients which lead to model failure.

6
2. Problem with temperature over steep orography
NH run at T159L60, ?t1h, NSITER1 Temperature
at level 59 after 27h
7
NH with NSITER2, ?t1h
Hydrostatic run, ?t1h
(with physics)
8
NH with NSITER1, ?t1h
Hydrostat. with no phys, ?t1h
(no physics)
9
Hydrostat. with no phys, ?t1h
NH with NSITER1, ?t1h
LVERAVE_HLUVFALSE
(no physics)
10
2. Problem with temperature over steep orography
NH run at T159L91, NSITER1,
NEPHYS3, Temperature at level 90 after 168h
?t0.5h
?t1h
11
Summary near surface sensitivities in the
vicinity of steep orography

• Decrease the time-step
• Switching off the averaging of surface winds
• Change the number of iterations
• Sensitivity to pointwise large surface wind
  accelerations (implicit convection scheme
  formulation)

LVERAVE_HLUVF
NSITER2 works best!
12
Physics Dynamics coupling

• Development of 2 options to call physics
• Call physics only in the last corrector step of
  the ICI scheme (adiabatic predictor steps)
  (NEPHYS3), NH(3)
• Call physics in the predictor step and use these
  physical tendencies in each subsequent corrector
  step (NEPHYS2), NH(2)

13
NH-IFS with physics

• Does the model climate of the NH version of IFS
  differ from the climate of the hydrostatic
  version for a Tl159L91 resolution ?

• 3 member ensemble
• 1 year
• daily SST forcing

14
zonal-mean zonal wind
NH(2)-H(NH)
15
zonal-mean zonal temperature
H(IFS)-ERA40
NH(2)-H(NH)
16
NH
H(NH)
diff
17
NH
H(NH)
H
500hPa Z, Northern Hemisphere
500hPa Z, Southern Hemisphere
500hPa T, Tropics
Anomaly correlation and rms error 12 cases,
Tl159L91
CPU time factor H1 H(NH)1.5 NH2 H(NH) and
NH used NSITER2
18
Physics Dynamics coupling

• 2 separate calls to vertical diffusion scheme,
  before and after calls to cloud and convection to
  test possibility for a better near surface
  balance in the last corrector step of
  nonhydrostatic modelhydrostatic test

19
Physics Dynamics coupling
1 call to vdfmain
2 calls to vdfmain
20
Physics-Dynamics couplingVertical diffusion
Anton Beljaars

• Negative tracer concentrations noticed despite a
  quasi-monotone advection scheme

21
Physics-Dynamics couplingVertical diffusion
(Kalnay and Kanamitsu, 1988)
Single-layer problem
22
Physics-Dynamics couplingVertical diffusion
Two-layer problem
depends on ? !!!
23
(DP)t
(DP)t?t
? 1.5
Dt?t
? 1
Anton Beljaars
24
Idealized flow past a mountain on the sphere

• Initial zonal flow, isothermal atmosphere, no
  physics
• Hydrostatic mountain NL/U gtgt1
• Non-hydrostatic mountain NL/U 1
• Froude number Nh/U 1

25
Hydrostatic model
Tl799L91, NL/U 900
Hydrostatic regime
Non-hydrostatic model
Horizontal divergence D
26
Tl159L91, NL/U 2.5
Near non-hydrostatic regime
Horizontal divergence D
hydrostatic model
non-hydrostatic model
27
noise
28
(No Transcript)
29
Next steps

• Further investigate the physics-dynamics
  interaction with NH-IFS
• Test NH-IFS in higher resolution
• Idealized tests in the NH regime on the sphere
• Improve the scores

30
dt3600s, p-p_hyd
dt3600s, horiz. divergence
dt1800s, p-p_hyd
dt1800s, horiz. divergence