fast currents in my roms model

[ckbenjamin]

I am currently having issues where a certain shallow region(~100m depth) of my domain is experiencing fast currents. The surface currents are ~100 cm/s and the bottom currents are ~50 cm/s. Some observational data I have suggest the currents should be in the magnitude of 1-10 cm/s.

This shallow gulf spans an area of ~5x5 degrees, and ~15 degrees away from the boundaries.
My model is at 6km horizontal resolution with 35 sigma levels.

Some parameters that I think may be relevant are provided below.
Anyone has any idea why the currents are so fast and how I may reduce them?

Thanks!

------------------------------------------------------------------------

THETA_S == 8.0d0
THETA_B == 0.4d0
TCLINE == 200.0d0
Vtransform == 2
Vstretching == 2
AKV_BAK == 1.0d-5
ad_AKV_fac == 1.0d0
AKK_BAK == 5.0d-6
AKP_BAK == 5.0d-6
TKENU2 == 0.0d0
TKENU4 == 0.0d0
GLS_P = 2.0d0
GLS_M = 1.0d0
GLS_N = -0.67d0
GLS_Kmin = 1.0d-8
GLS_Pmin = 1.0d-8
GLS_CMU0 = 0.5544d0
GLS_C1 = 1.00d0
GLS_C2 = 1.22d0
GLS_C3M = 0.1d0
GLS_C3P = 1.0d0
GLS_SIGK = 0.8d0
GLS_SIGP = 1.07d0


#define GLS_MIXING
#undef CANUTO_A
#undef CANUTO_B
#undef CHARNOK
#undef CRAIG_BANNER
#undef KANTHA_CLAYSON
#undef K_C2ADVECTION
#undef K_C4ADVECTION
#define N2S2_HORAVG
#undef ZOS_HSIG
#undef TKE_WAVEDISS
#define BULK_FLUXES

[m.hadfield]

I believe you can get ROMS to output the surface stresses, which in your case should have been calculated by BULK_FLUXES. Do these look reasonable?

[rduran]

with such dx,dsigma and total depth it is possible you bathymetry would need much smoothing, do you have a slope? how do your r-factors look? a false pressure gradient certainly produces currents.

[ckbenjamin]

Mark: I have not paid much attention to the surface stresses. I do have them specified in my ocean.in, but its not produced. Should I define anything in cpp?

Rodrigo: I do have slopes around, but I did not think too much of them as the max depth is about 70m hence the gradient is not that extreme compared to deep oceans. I have rx0, 0.3, rx1, 20. If you think bathymetry may be the main culprit here, perhaps I could try to diagnose with an experiment where I artificially flatten off the bottom of this basin?

[m.hadfield]

An rx1 of 20 is too high. Somewhere in the vicinity of 5 to 8 should be OK.

In my experience, the failure mode for large rx1 is that a large patch of the bottom layer decides, for no obvious reason, to move rapidly in one direction or other. This produces large velocities at the coastline, which crashes the model. Because the failure occurs very rapidly, and the root cause is distant from the location where the CFL criterion is exceeded, it can be hard to diagnose. (I speak from experience!)

The cure is to increase the thickness of the bottom model layer, by reducing N, or reducing the degree of near-bottom refinement. Reducing the bottom slope may help, too, but the degree of smoothing that's required to make a significant difference could be unacceptable.

[ckbenjamin]

I do have a recent run, diagnosing if it is due to the high rx1. In this run, I tried reducing N and got a rx1 of 7.32. However, the issue of fast currents both at the surface and bottom persisted.

[rduran]

the proper diagnostic test is to run your model with zero everything (ie no forcing no initial velocities) and only a horizontally-constant vertical (stable) profile of density. perhaps constant salinity and a typical temperature profile will do if temperature controls density in your region. any velocities produced there will be due to pressure gradient error, you will get an estimate of their magnitude.
closed boundaries is a possibility but after some time of integration beware of waves generated by pressure gradient error bouncing back into the domain and adding to the error.

a shallow sea is actually a double hit: dsigma is smaller than in a deeper ocean and when you divide by depth you get a bigger number, both contribute making it more difficult for hydrostatic consistency to be satisfied. and the bigger your horizontal spacing the worse this matter gets. In summary you should do 1) the test mentioned above and 2) not expect numerical convergence unless you improve your hydrostatic consistency (probably good idea to look it up). Bathymetry slope is the last parameter in hydrostatic consistency, the smaller the better but in total there are 4 quantities to look after as mentioned above.

[ckbenjamin]

Thanks! I will look it up.

[ckbenjamin]

Update:
The diagnostic run with zero everything yielded almost zero-everything-currents at the top and bottom in this area of concern.

I had a separate realistic run where I had realistic everything, but reduced the wind forcings by a factor of 10 and the surface/bottom currents dropped tremendously.

Would this imply there is excessive momentum transfer from the winds? and subsequently down the water column?

[ckbenjamin]

To update this thread, I have gotten much better results after paying more attention to my bottom stresses.

[rduran]

thanks for sharing! it's always good to earn in experience