[MITgcm-support] Advection and turbulence schemes in realistic NH simulation

Andrea Cimatoribus andrea.cimatoribus at epfl.ch
Tue Jul 4 03:36:43 EDT 2017


Dear all,
I am currently running a 113m (horizontal) resolution configuration, 
with non-hydrostatic dynamics included, mostly because I am considering 
complex topography with steep slopes.

I have been scanning the literature in search for similar 
configurations, but I have not found much. It seems that most of the 
work at high-res/NH is in idealised conditions and periodic domains 
(HAPPY TO BE CORRECTED).

Finding myself in uncharted territory, I have two questions:

1) Concerning the avection scheme
Also following a ~recent discussion, I am using:

  usejamartwetpoints=.TRUE.,
  vectorInvariantMomentum=.TRUE.,
  highOrderVorticity=.TRUE.,

which seems to be working fine (no noise, stable). However, I recently 
noticed another option in my STDOUT:

(PID.TID 0000.0001) selectVortScheme= /* V.I Scheme selector for 
Vorticity-Term */
(PID.TID 0000.0001)                       0
(PID.TID 0000.0001)    = 0 : enstrophy (Shallow-Water Eq.) conserving 
scheme by Sadourny, JAS 75
(PID.TID 0000.0001)    = 1 : same as 0 with modified hFac
(PID.TID 0000.0001)    = 2 : energy conserving scheme (used by Sadourny 
in JAS 75 paper)
(PID.TID 0000.0001)    = 3 : energy (general) and enstrophy (2D, 
nonDiv.) conserving scheme
(PID.TID 0000.0001)          from Sadourny (Burridge & Haseler, ECMWF 
Rep.4, 1977)

I couldn't find any specific documentation on this, but it seems that 
some in the ML rather use selectVortScheme=2. Does anyone have any 
further insight? I am planning to look in some detail at the dynamics 
near the boundaries, so I would like to be sure about my vorticity balances.

2) Concerning the turbulence model
As far as I understand it, strictly speaking, KPP or GGL90 (my current 
choice) are not meant for non-hydrostatic. This due to both technical 
(e.g. implicit viscosity, shear in GGL90 from hydrostatic part alone) 
and fundamental reasons (e.g. risk of "counting overturns" twice). 
However, GGL90 does work in my configuration, and results are not at all 
bad. BTW, I am using Prather advection scheme and use molecular values 
for background diffusivities. Is there any alternative to my very 
pragmatic current approach?

Thank you very much for your feedback,
Andrea



-- 
Andrea Cimatoribus
postdoctoral researcher
EPFL ENAC IIE ECOL
https://people.epfl.ch/andrea.cimatoribus



More information about the MITgcm-support mailing list