computation of zeta for stable buoyancy forcing and wind stress about cvmix-src HOT 6 CLOSED

This change is documented in the 2006 Danabasoglu, et al., paper Diurnal Coupling in the Tropical Oceans of CCSM3 (found in Journal of Climate). From Appendix A (page 2361):

The turbulent velocity scale, ws, usually equals the friction velocity at the surface and increases in unstable forcing over the uppermost 10% of the boundary layer, then is held constant below in order to avoid excessively large values as depth increases. For consistency, a similar constraint is now applied in stable forcing that limits the decrease in ws with depth, and hence increases Vt.

from cvmix-src.

Luke,

I am having a tough time seeing a significant difference between your two
KPP runs. It would help to have the three figures together so that I can
better compare. But from my initial look, it is tough to say that your
modified KPP run is closer to LES.

Perhaps I am not seeing something that you are focusing on...?

Steve

On Wed, Oct 14, 2015 at 11:24 AM, Michael Levy [email protected]
wrote:

This change is documented in the 2006 Danabasoglu, et al., paper Diurnal
Coupling in the Tropical Oceans of CCSM3
http://journals.ametsoc.org/doi/abs/10.1175/JCLI3739.1 (found in
Journal of Climate). From Appendix A (page 2361):

The turbulent velocity scale, ws, usually equals the friction velocity at
the surface and increases in unstable forcing over the uppermost 10% of the
boundary layer, then is held constant below in order to avoid excessively
large values as depth increases. For consistency, a similar constraint is
now applied in stable forcing that limits the decrease in ws with depth,
and hence increases Vt.

—
Reply to this email directly or view it on GitHub
#56 (comment).

Dr. Stephen M. Griffies
NOAA Geophysical Fluid Dynamics Lab
201 Forrestal Road
Princeton, NJ 08542
USA

from cvmix-src.

Steve,
It is certainly not a massive difference, but it is there. I've annotated the images from the two runs and placed them closer together below. Perhaps this will help. I've added two vertical lines at the same times (1.5 days and 3.0 days). What I'm noticing is that at day 1.5, the reddish contour has already outcropped/reached the surface in the CVMIX formulation from Danabasoglu et al 2006. When the turbulent scale is not limited, the temperature stays warmer longer. This is seen in the LES. If you examine day 1.5, the temperature structure matches better with my test case (altering CVMIX). If you then look at day 3, the dark red temperature surface has nearly reached the surface in the current CVMIX form, but has not in the test case. Again, this result is closer to what is observed in LES. As alluded to in Danabasoglu et al 2006, the artificial limitation of sigma deepens the boundary layer depth minimally (approximately 0.5 meters difference in the two cases by the end of 3.5 days)

Run with sigma limitation (current cvmix)

without the limitation on sigma

Finally LES

In essence, it appears to me that the temperature structure observed with the cvmix changed is closer to the LES result (horizontally averaged temperature profiles). If this is still clear as mud to others besides myself, I can go back and form plots of theta_original_cvmix - theta_LES and theta_changed_cvmix - theta_LES.

In general, however, I'm not totally comfortable with, or perhaps not understanding, the reasoning for making the change outlined Danabasoglu et al 2006. By my reading, it seems that the turbulent velocity scaling was changed to make CCSM3 boundary layer depths deeper in the tropical simulations shown in the paper (closer to observations). But, there is really no physical rationale behind this change. Then again, there really is no physical rationale to changing back to the original LMD94 formulation. One of my current tasks is to develop an analysis to help visualize the turbulent velocity profile in the boundary layer in my corresponding LES result. I will post this figure here as soon as I get it made.

from cvmix-src.

Hi,

I now see your point. Thanks for the clarification.

One suggestion is to code an option to facilitate further testing of your
modified approach. Can you fork a branch that provides the code mod that
allows for this change to be toggled, with the default set to the current
settings? Please document your code mod so that folks can know what is at
question. Be precise, in that you should point to the Danabasoglu et al
(2006) equations/discussion that you are questioning.

Please be sure to allow for present answers to go unchanged, as that will
facilitate easier testing by others.

When you are happy with your fork, then let others know so that they can
test the change. If others concur, we may wish to incorporate this to the
main branch. But I defer to the NCAR folks to make that decision...

Thanks,
Steve

On Wed, Oct 14, 2015 at 2:10 PM, Luke Van Roekel [email protected]
wrote:

Steve,
It is certainly not a massive difference, but it is there. I've annotated
the images from the two runs and placed them closer together below. Perhaps
this will help. I've added two vertical lines at the same times (1.5 days
and 3.0 days). What I'm noticing is that at day 1.5, the reddish contour
has already outcropped/reached the surface in the CVMIX formulation from
Danabasoglu et al 2006. When the turbulent scale is not limited, the
temperature stays warmer longer. This is seen in the LES. If you examine
day 1.5, the temperature structure matches better with my test case
(altering CVMIX). If you then look at day 3, the dark red temperature
surface has nearly reached the surface in the current CVMIX form, but has
not in the test case. Again, this result is closer to what is observed in
LES. As alluded to in Danabasoglu et al 2006, the artificial limitation of
sigma deepens the boundary layer depth minimally (approximately 0.5 meters
difference in the two cases by the end of 3.5 days)

Run with sigma limitation (current cvmix)

[image: mpas_shear_fix_wind_stable_buoy]
https://cloud.githubusercontent.com/assets/13662545/10492837/48455918-726b-11e5-89be-b09385f992e6.png

without the limitation on sigma

[image: cvmix_turb_scale_fix_wind_stable_buoy]
https://cloud.githubusercontent.com/assets/13662545/10492914/99bb63f0-726b-11e5-97df-93e707c83b4b.png

Finally LES

[image: les_wind_stable_buoy]
https://cloud.githubusercontent.com/assets/13662545/10492971/d7aabe54-726b-11e5-8bac-1168966c37dd.png

In essence, it appears to me that the temperature structure observed with
the cvmix changed is closer to the LES result (horizontally averaged
temperature profiles). If this is still clear as mud to others besides
myself, I can go back and form plots of theta_original_cvmix - theta_LES
and theta_changed_cvmix - theta_LES.

In general, however, I'm not totally comfortable with, or perhaps not
understanding, the reasoning for making the change outlined Danabasoglu et
al 2006. By my reading, it seems that the turbulent velocity scaling was
changed to make CCSM3 boundary layer depths deeper in the tropical
simulations shown in the paper (closer to observations). But, there is
really no physical rationale behind this change. Then again, there really
is no physical rationale to changing back to the original LMD94
formulation. One of my current tasks is to develop an analysis to help
visualize the turbulent velocity profile in the boundary layer in my
corresponding LES result. I will post this figure here as soon as I get it
made.

—
Reply to this email directly or view it on GitHub
#56 (comment).

Dr. Stephen M. Griffies
NOAA Geophysical Fluid Dynamics Lab
201 Forrestal Road
Princeton, NJ 08542
USA

from cvmix-src.

Steve (et al.),
I have created a fork and put in a pull request (#57). I was not sure if this was where you wanted me to reference the fork for testing. The relevant text and equations from Large et al 1994 and Danabasoglu et al 2006 are included in the PR description. I will also continue testing this branch in my SCM version of MPAS-O relative to LES.

Regards,
Luke

from cvmix-src.

Fixed via #62 on March 7th

from cvmix-src.