way of reducing vertical mixing ...

[turuncu]

Hi,

I had a problem with ROMS. The vertical mixing seems too much and i could not generate cold intermadiate layer (CIL) in my domain (Black Sea). I heard that sigma layer models has a problem in vertical mixing when there is a sharp gradient in the bathymetry. So, i ma just looking for a solution to reduce the vertical mixing. i searched internet and i found following information

http://woodshole.er.usgs.gov/operations ... ode21.html

it basically suggest to remove domain averaged T, S and rho profiles just before calculating horizontal differences. Does this help to fix this vertical over mixing problem? I just wonder that does it implemented in the ROMS? What is the common practice to overcome this problem. I did lots of test runs with different coefficients and parameterizations but the results similar. There is no strong CIL. Anyway, any suggestion could be helpful.

Regards,

--ufuk

[kate]

We used to remove functions of z only - twenty years ago. As that site points out, the larger the domain, the more troublesome this is.

Things have changed in the ROMS sigma-coordinate in that time. The pressure gradient errors are smaller, we have the rotated horizontal mixing options, and you can try the split advection scheme. If you get really desperate, you can try the ROMS AGRIF branch where some of these things are even better, such as a stiffened equation of state.

[arango]

The turbulent schemes like MY-2.5 or GLS tend to give high vertical mixing coefficients if you have strong winds and buoyancy forcing. Try the KPP scheme to see what you get. The mixing is more controlled in LMD_MIXING.

[turuncu]

Thanks Kate and Arango. Yes, KPP helps but in this case the surface currents are not well represented by model. It was better with k-omega. I'll try to do more test. By the way, i also tried to split advection scheme but in that case it also activates biharmonic horizontal diffusion and i could not keep the model stable even if i change the TNU4 parameter. Any suggestion?