Problem with KPP

[nencio]

Hi all,
I'm trying to setup a numerical experiment for the Hawaii Island region.
The domain is 241x133 with a resolution of about 1/10 of degree.
You can check the CPP options I'm using and the values for the various parameters in the attached KPP_log.txt file.

KPP_log.txt

(19.82 KiB) Downloaded 327 times

Despite I can succesfully run the first year of spin-up (meaning that the model doesn't blow-up!! I start from rest and uniform conditions), after couple of months of simulation the velocity field in the first vertical levels becomes characterized by weird/unrealistic stripes:

This is the surface field of the V-momentum component after the third month of spin-up.
The figure was generated with ncview using the "detailed" colorbar.

kpp_surfV_dt91.gif (73.67 KiB) Viewed 4334 times

Playing around trying to figure out what was causing the problem, I ran a simulation with the exact same parameters as the previous one, but using the GLS as vertical closure scheme instead of the LMD. As you can see the velocity field looks ok in this case:

gls_surfV_dt91.gif (69.62 KiB) Viewed 4334 times

I then checked the depth of the surface boundary layer (Hsbl) for the simulation with the LMD scheme activated. That field, as well, was characterized by similar stripes as the ones present in the velocity field.

Assuming at this point that the origin of the problem is the LMD vertical closure scheme (and not being experienced at all with it) I would like to know if anybody had a similar problem with it, and most importantly if anybody has any suggestion on how I could fix it.


I looked at the examples that come with ROMS, and the ones that have a resolution similar to my experiment (DAMEE_4, SCB, NATL) all use the same CPP options (LMD_CONVEC, LMD_MIXING, LMD_NONLOCAL, LMD_RIMIX, LMD_SKPP) and the same values for the parameters in the .in file that I'm using.

Thanks a lot,
francesco

[kate]

I use the LMD_SHAPIRO option to smooth out hsbl.

[nencio]

Hi kate,
thanks for the quick reply. I tried your suggestion and the hslb field is much smoother now. However the velocity field still shows those weird stripes:

KPP_shap_surfV_dt91.gif (76.19 KiB) Viewed 4299 times

anything else I can try??
...and most importantly: anybody who had a similar problem??

thanks,
francesco

[kate]

Is there some reason not to use one of the GLS schemes? Actually, in looking at the v-velocity, I'd be more worried about your boundary conditions. What are you using? It looks like you have a western wall jet and other lovely boundary eddies.

[nencio]

don't have any particular reason not to use the GLS, other than it takes a little longer to run the simulation.
The thing is that I would just like to know, if the one I'm having with the KPP is actually a problem, and if it is, then what is causing it.

Regarding my boundary conditions: I'm using 4 walls at the moment (that's the reason why I got all those eddies along the boundaries of my domain). I've already run a simulation for the whole year with 4 open boundaries and radiation conditions, but since I was having that problem with the velocity field, I decided to simplify my simulation as much as possible in order to narrow down the possible sources of the problem.

- francesco

[cae]

With coastal simulations, we've experienced a nearshore checkerboard pattern using KPP during periods of weak wind stress. This pattern was reduced some with horizontal smoothing, but still was disturbing. We've had better luck from this perspective with gls. Not exactly the same issue that you have, although probably related.

Good luck!