Page 1© Crown copyright 2005 RF01/RF02: LES sensitivity studies Adrian Lock and Eoin Whelan.
-
date post
21-Dec-2015 -
Category
Documents
-
view
214 -
download
0
Transcript of Page 1© Crown copyright 2005 RF01/RF02: LES sensitivity studies Adrian Lock and Eoin Whelan.
© Crown copyright 2005 Page 2
Starting point: RF01 intercomparison
Met Office LES was low down the RF01 LWP league table!
And at the top for subgrid heat flux across the inversion
© Crown copyright 2005 Page 3
Starting point: RF01 intercomparison
Disappointing as the Met Office LES subgrid model should be stable for RF01:
Smagorinsky type + MacVean and Mason Ri For RF01, M&M should give Ri>0 implying little subgrid
mixing across the inversion
M&M, k=0.7 R&D, k=0.23
RF01, k~0.5
x
Ri < 0
e
tq
( / )e
p t
kL c q
© Crown copyright 2005 Page 4
RF01: simple changes
1. MacVean (1993): reduce neutral mixing length towards the inversion to reflect geometrical constraint on eddy size
2. Use monotone scheme (rather than 1st order upwind) for subsidence forcing
z=0
ziz
neutl
© Crown copyright 2005 Page 5
Impact in RF01
These ‘simple’ changes give some increase in LWP But still a factor of 2 too low
© Crown copyright 2005 Page 6
Bjorn’s fix
Switch off subgrid model for scalars (above 750m)Crude but effective – same as Bjorn
Switch off subgrid model completely (above 750m)Disaster - ?same as Bjorn?
© Crown copyright 2005 Page 7
Monotone advection of all variables
Monotone advection of momentum (as well as scalars) gives results almost identical to Bjorn’s fixCan get to the top of the LWP league table without having to half switch off the subgrid model!
© Crown copyright 2005 Page 8
Is the diffusion implicit with monotone advection equivalent to having a more active subgrid model? Try standard (non-monotone) advection but with cs=0.32
(instead of 0.23, ) Some improvement in LWP but still some way short Could increase cs further but subgrid fluxes and entrainment
would increase further
Increase subgrid diffusion
0 sc x
© Crown copyright 2005 Page 9
Turbulence or noise?
Bjorn’s fix needs the subgrid model on for momentum – why? We (all?) use monotone advection schemes for scalars, why not
for momentum? W field doesn’t look too noisy Monotone advection gives loss of energy at smaller scales…
Monotone advection of momentumCentred-difference advection of momentum
© Crown copyright 2005 Page 10
Spectra
Monotone advection of momentum leads to reduction in energy at scales close to the grid-scale Similar to cs=0.32
w spectra at 500m
© Crown copyright 2005 Page 11
Spectra at entrainment flux level
Very different spectra between monotone and non-monotone advection of momentum just below the inversion
Is the extra energy near the grid scale with non-monotone advection just numerical noise?
w spectra at ziv spectra at zi
© Crown copyright 2005 Page 12
Turbulence or noise?
Horizontal momentum field certainly looks noisy at the inversion
Why should we believe this noise any more than the noise we don’t like in the scalar fields?
Spurious noise in the momentum fields would reduce Ri(>0) and so increase subgrid scalar mixing across the stably stratified inversion (except with Bjorn’s fix)
Centred-difference advection of momentum Monotone advection of momentum
© Crown copyright 2005 Page 13
Monotone advection of all variables
No problem with monotone advection of momentum in matching the observed w’w’
© Crown copyright 2005 Page 14
TKE budget
Total dissipation also very similar between Bjorn’s fix and monotone momentum:
Total TKE dissipation
= subgrid model +Advection scheme
(=residual)
Bjorn’s fix
Monotone momentum
© Crown copyright 2005 Page 15
Resolution sensitivity
Run with: coarse (Dz=10m,Dx=35m), standard (Dz=5m,Dx=35m), fine (Dz=2.5m, Dx=17.5m)
Monotone advection ~ centred differences at doubled resolution Not really converged
Higher resolution gives reduced entrainment and thence increased LWP Only monotone fine resolution has cloud base ~ constant, as observed
© Crown copyright 2005 Page 17
Conclusions
Spurious noise in momentum fields close to the inversion can enhance subgrid fluxes there leading to excessive entrainment at standard resolutions (Dz=5m, Dx=35-50m)
Met Office LES gives apparently realistic results when monotone advection is used on all variables Results are equivalent to a centred-difference scheme with
double the resolution Avoids the need for Bjorn’s fix But simulations have not converged by Dz=2.5m, Dx=17.5m