COMODO, Evaluation of numerical methods for Ocean Modeling

COMODO, Evaluation of
numerical methods for Ocean
Modeling
Yves Soufflet, Florent Lyard, Laurent Debreu, Patrick
Marchesielo ...and more
LEGOS, Toulouse
IMUM 2014, Lisbon
http://www.comodo-ocean.fr
1/20
The Comodo Project
7 labs
7 tasks
• Dissipation mechanisms
• Momentum advection
• Ifremer, LPO, LEGOS, LEGI,
INRIA, SHOM, LOCEAN, LA
schemes
• Time integrators
6 Models
• Tracers advection-diffusion
MARS, NEMO, ROMS,
Symphonie, HYCOM, TUGO
• Vertical coordinate systems
• Unstructured mesh methods
• Non hydrostatic
http://www.comodo-ocean.fr
1 common tool: pycomodo
library (Marc Honnorat INRIA)
2/20
The Comodo Project
http://www.comodo-ocean.fr
3/20
The Comodo Project
http://www.comodo-ocean.fr
4/20
7 Testcases
• Baroclinic Vortex
Momentum scheme and dispersion
• Eddy topography iteraction
Topographic rectification, PV and enstrophy conservation
• Thacker
bbl and adiabatic flow along topography
• Upwelling
bottom flow and vertical coordinates
• Internal Wave
• Lock Exchange
spurious diapycnal diffusion
• Baroclinic Jet
Effective model resolution
http://www.comodo-ocean.fr
5/20
Internal Wave
Objectives
configuration
• Validate the pressure
gradient calculation,
• Sensitivity of the generation
and propagation of IW,
• Investigate the effective
resolution in relationship to
horizontal and vertical
resolution,
• Address the extension
toward non-hydrostatic
methods,
http://www.comodo-ocean.fr
• Test outgoing OBC for
baroclinic fields in
hydrostatic and
non-hydrostatic dynamics
6/20
Internal Wave with Tugo
http://www.comodo-ocean.fr
7/20
Lock Exchange
Objectives
configuration
• Quantification of the
spurious diapycnal mixing
• Investigate the accuracy of
advection schemes
• Impact of numerical
choices on numerical
solution
• Setup following Haidvogel
and Beckman, Burchard
and Rennau 2008
• Diagnostics based on Illicak
(OM 2012), RPE , density
class and front location
http://www.comodo-ocean.fr
8/20
Lock Exchange in Fluidity-ICOM
http://www.comodo-ocean.fr
9/20
Baroclinic Jet: Surface Vorticity and KE
Spectrum
http://www.comodo-ocean.fr
10/20
ROMS NEMO comparison, Surface
Vorticity
Figure: Colorbar symetric from -0.00015 to 0.00015
http://www.comodo-ocean.fr
11/20
Profiles
Figure: 20 years average profiles for ROMS (plain line) and NEMO (dashed)
http://www.comodo-ocean.fr
12/20
ROMS NEMO comparison: KE spectra
Figure: 20 years average spectra at 10m depth for ROMS (plain line) and
NEMO (dashed)
http://www.comodo-ocean.fr
13/20
Injection <w’b’> at different resolution
Figure: Injection Spectra at 10m
http://www.comodo-ocean.fr
14/20
Diffusion at different resolution
• Amount of dissipation decrease with resolution
• Dissipation range is also reduced towards small scale
http://www.comodo-ocean.fr
15/20
Transfert at different resolution
• Direct cascade in the dissipative range
http://www.comodo-ocean.fr
16/20
Barotropic/baroclinic coupling filter
• Effect of temporal filter: spectra for 5 Km resolution at 10m
• No effect at small scales
http://www.comodo-ocean.fr
17/20
Effect of Diapycnal mixing
• ROMS run with RSUP3-ISO: Rotation of diffusion tensor
• Extension of the injection range at small scales
http://www.comodo-ocean.fr
18/20
Profiles at 5km
• Large effect of diapycnal mixing at the surface
• Preserving the stratification allows more APE to be converted
http://www.comodo-ocean.fr
19/20
Final word
The baroclinic Jet
• Spectral budget help to
identify sources of
dissipation
Other test cases
• Looking for volunteers...
• Which test case can YOU
do?
• Difficult to point exactly to
the source of differences
between model
• Clear/important role of the
diffusion of tracer
http://www.comodo-ocean.fr
20/20
`