Noisy SST

[xavierC]

Hello,
I am doing some ROMS modelling and although it goes quite smooth, i have some noise on the temperature field, mainly at the surface, as you can see on the figure attached.

I am using
#define UV_ADV
#define TS_U3ADV_SPLIT
advections schemes (but i have tried others without any success)

and i have been playing around with viscosity and difusivity.
/*Horizontal Mixing */
#define TS_DIF4
#define TS_DIF2
#define UV_VIS2
/*Vertical Mixing*/
#define MIX_GEO_TS
#define MIX_GEO_UV
#define LMD_MIXING
#define LMD_BKPP
#define LMD_CONVEC
#define LMD_RIMIX
#define LMD_SKPP
#undef LIMIT_VDIFF
#undef LIMIT_VVISC

The model is stable and all other field are (or at least seem) smooth.

Would you have any idea of what i should do to resolve this issu and get smooth temperature ?

I forgot to mentionne that my time step is 100s and horizontal resolution 2.5km. Although i think it already too small i tried to reduce it, but once again without ay success .

Thanks in advance
Xavier

[c.drinkorn]

Hi Xavier,

I switched from MIX_GEO_TS to MIX_S_TS in order to solve a similar problem at the bottom. Maybe it also helps you.

//Cate

[xavierC]

Hi Xavier,

I switched from MIX_GEO_TS to MIX_S_TS in order to solve a similar problem at the bottom. Maybe it also helps you.

//Cate

Thanks a lot for your answer Cate. I just made a try, but no sucess once again...

[c.drinkorn]

Then the only idea I have is to enable TS_MPDATA which will "force" the fields to be in 3d consistency...

[jcwarner]

do the atm forcing fields look smooth ?
you could try mixing from my25 or gls and then activate N2S2_HORAVG.

-j

[kate]

The equivalent horizontal smoothing for LMD is LMD_SHAPIRO, to horizontally filter the mixed layer depth. I've heard from some other colleagues that they too need to smooth the mixed layer depth to avoid that sort of noise.

As for MIX_S_TS, it's a really bad idea over a sloping bottom and can lead to unphysical cross-isopycnal mixing.

[c.drinkorn]

Code: Select all

As for MIX_S_TS, it's a really bad idea over a sloping bottom and can lead to unphysical cross-isopycnal mixing.

I hear you! But there was no way to solve the issue of punctual alternating high/low temperatures at the slope when mixing along z-planes which would even cause the model to blow up. The model doesn't "know" the sigma levels are aligned to the slope. It just detects strong gradients between adjacent cells and calculates way too high advection velocities leading to a sort of snowball effect, I think. How can this be solved otherwise?

[xavierC]

Hello,
thanks for your answsers, still have to try GLS or MY2.5, but I managed to try the "easy one" suggested by kate and i got a strong improvments of my solution. With some presistent noise only very near coast.

[kate]

Code: Select all

As for MIX_S_TS, it's a really bad idea over a sloping bottom and can lead to unphysical cross-isopycnal mixing.

I hear you! But there was no way to solve the issue of punctual alternating high/low temperatures at the slope when mixing along z-planes which would even cause the model to blow up. The model doesn't "know" the sigma levels are aligned to the slope. It just detects strong gradients between adjacent cells and calculates way too high advection velocities leading to a sort of snowball effect, I think. How can this be solved otherwise?

Maybe it depends on your priorities - what are the time and space scales you'd like to do right? Can you get a smooth enough result with the third-order advection and no diffusivity? Do you need the split advection scheme which may or may not be working well enough in the ROMS you are using? Do you need a different ROMS or another model entirely? Or can you live with what your ROMS is giving you?

[beaufort]

Just wanted to add that this sounds like an issue I've been having with noisy SST (and SSS), particularly in regions of low wind stress - albeit in CROCO v1.1 rather than ROMS - which I solved by switching mixing from LMD to GLS.