Numerical Noise in distant regions by changing runoff

Versions
1

Hi everyone,

I’m conducting a regional parametric study where I perturbed a river and observed unexpected noise in a distant region, just after few hours (very low) but starts to grow as days pass and becomes significantly high, like after few months i saw a change in Atlantic after perturbating a river in Eastern Mediterranean, i was able to see a change of around 1 psu in atlantic thats not physically advected. The change is in tracers, velocities, and sea surface height (SSH). The change may originated near the bathymetry since i saw change in velocities and tracers near the base like in the deep ocean in Atlantic

Here are the key details of my setup:
I am using a 1/16° grid with z-coordinates and partial steps (no AGRIF zoom) with Split-Explicit formulation.
I am using open boundary only in the Atlantic, with flow relaxation for tracers and baroclinic variables, and Flather conditions for barotropic variables and SSH.
For time stepping, my Baroclinic step = 200 seconds; barotropic step = ~2 seconds using nn_baro (also tested CFL-dependent barotropic stepping).
I am using MUSCL for tracers advection. I am using vector form for momentum advection however i also tested flux form.
My river is just at the top cell with runoff and salinity only.
I tried changing turbulence schemes, the use of GLS/TKE closure models reduced the error signficantly in comparison with Richardson number-dependent mixing however smaller error persist that grows over months. I am using MPI, and in NEMO 5 i used halo cells = 2.
I ran experiments using both NEMO 3.6 and NEMO 5 beta, testing both linear and non-linear SSH formulations. I also tried different configurations, including:
• Reducing the timestep (affected but didn’t eliminate the noise)
• Splitting the river into multiple cells instead of a single input.
• Using the Runge-Kutta scheme in NEMO 5.
• Running the simulation on a smaller domain with finer resolution (similar kindof noise observed).

I would greatly appreciate any insights or suggestions on how to address this issue. Please let me know if i could provide any more information that could be helpful.

2

Hello Saad,

Welcome to Nemo. Could you be a bit more precise regarding the perturbation you add to the river runoff ?

Cheers,

Robinson

3

Dear Robison, thank you for reading. I am running two identical cases and i just change the monthly runoff of the river (two different runoffs with different seasonal cycle). Apart from the flow change around the grid cell of the river, i see change in far away regions.

4

And are you sure the distant noise appears only if you change the runoff ? Could you please send me the runoff files before & after (you can gzip them) to robinson.hordoir@hi.no

5

Yes, thats the only thing that i change. I will share the files with you. Thank you!