rules of thumb for key parameters

[jul10lara]

Hello everybody!

I use ROMSTOOLS and I'm trying to configure my own simulation for a specific area of the Pacific Ocean, so I need to specify the values of certain parameters, the matter is that I have no idea how to select those values. For example, in the ROMSTOOLS User's Guide there is a rule of thumb to properly specify hmin according to dl (the rule is: dl=1 --> hmin=300; dl=1/2 --> hmin=150 ...)

Is there a rule of thumb, or any reference values to properly configure the next parameters??

In the romstools_param.m file:

- N (number of vertical levels)
- theta_s
- theta_b
- hc
- hmax_coast
- rtarget
- n_filter_deep_topo
- n_filter_final
- zref (level of reference for geostrophy calculation)

In the roms.in files:

- rho0
- lateral_visc: VISC2, VISC4 [m²/sec]
- tracer_diff2: TNU2(1:NT) [m²/sec]
- tracer_fiff4: TNU4(1:NT) [m^4/sec]
- vertical_mixing: Akv_bak, Akt_bak [m²/sec]
- bottom_drag : RDRG (m/s), RDRG2, Zob (m). Cdb_min, Cdb_max
- gamma2
- sponge: X_SPONGE (m), V_SPONGE (m²/sec)
- nudg_cof: TauT_in, TauT_out, TauM_in, TauM_out (days for all)

For nesting:

- n-band

Thanks every body. I hope you can help me!!!!

[milena]

Hi there!

One thing I can tell you for sure is don't mess too much with theta_s, theta_b and hc, as the grid you get can become very unstable. Once you found a set of parameters that works for you, keep it! I had luck with theta_s=5, theta_b=0 or 0.4 (for a South Atlantic and Eastern Pacific application, respectively) and hc=10. Other people may have different suggestions. rtarget is similar in this sense (although more subtle: the model won't likely blow out right away because of rtarget); you may wish to stay conservative, like choosing 0.2 or 0.25.
In my experience, the grid parameters are the most important in ROMS.

N really depends on how much resolution you want in the mixed layer or at the coast. People use 30, 40,60. You may want to fix it, then plot the sigma levels and see if you are satisfied with your vertical resolution at various depths. It also depends on how large a grid you can afford, of course.

Here are the other choices I made for the remaining parameters:
- rho0 = 1025
- lateral_visc: VISC2 = 4 or 10 (for grid resolutions of 0.1deg, 0.25deg respectively)
- tracer_diff2: TNU2(1:NT) = same as VISC2
- vertical_mixing: Akv_bak, Akt_bak [m²/sec] = 1.0d-5 (mind you: these are not needed if you use KPP)
- gamma2 = 1 (free slip bc's)
- sponge: X_SPONGE (m), V_SPONGE (m²/sec) = 150e3, 100 (typical mesoscale eddy scale~150km)
- nudg_cof: TauT_in, TauT_out, TauM_in, TauM_out = 1, 360, 1, 360 (you could do endless experiments with these..)
- hmax_coast = 200 or 500
- n_filter_deep_topo = 4 (default)
- n_filter_final = 2 (default)
- zref (default)
- nband = I chose as many points as would fit within ~0.5-1deg, according to the child resolution (for example, nband=12 for dx=1/12deg)

Hope it helps!
my best,
Milena

[jul10lara]

Milena, thank you very much for the reply, it will be very useful!!!

I found that 1<theta_s<20 and 0<theta_b<1
Also, I was told that a proper thickness for the sponge could be the equivalent to 10 cells,

I hope someone more, can give me (us) other tips!!!

Than you in advance

Best regards.

[milena]

I found that 1<theta_s<20 and 0<theta_b<1

Yes, but that is a huge range of values.
For example, I first chose thetas=6, thetab=0, hc=20 (I wanted more levels in the surface mixed layer), and the grid was VERY unstable (even reducing the time step several times wouldn't help). Just by going to thetas=5 and hc=10 changed everything (same surface forcing, bc fields, and other parameters as before, but this configuration just ran without problems with a decent time step).

Also, I was told that a proper thickness for the sponge could be the equivalent to 10 cells

Personally, I think the sponge has only the purpose to kill stubborn eddies that get stuck near the boundaries, especially in corners (would be nice not to use a sponge at all..). So, the sponge layer should be at least as wide as a typical mesoscale eddy, ~150km (different number of cells according to your resolution).

best,
Milena

[jul10lara]

Dear Milena,

thanks for your tips
I'll keep them in mind!!!!

Regards