Ocean is too warm in 1 deg Tripolar simulation

[taimoorsohail]

The ocean has warmed up too much after 25 years:

[ Info: time: 10999.583 days, iteration: 805650, Δt: 20 minutes, max|u|: (2.72e+00, 2.76e+00, 9.82e-03) m s⁻¹, extrema(T): (80.44, -2.17) ᵒC, wall time: 1.504 seconds 
[ Info: time: 10999.722 days, iteration: 805660, Δt: 20 minutes, max|u|: (2.72e+00, 2.76e+00, 9.82e-03) m s⁻¹, extrema(T): (81.09, -2.17) ᵒC, wall time: 1.482 seconds 
[ Info: time: 10999.861 days, iteration: 805670, Δt: 20 minutes, max|u|: (2.72e+00, 2.77e+00, 9.85e-03) m s⁻¹, extrema(T): (78.27, -2.16) ᵒC, wall time: 1.481 seconds 

I am looking into the cause of this, particularly the surface forcing.

[navidcy]

@simone-silvestri that's a slightly modified version of the 1 degree model from ClimaOcean.
Have you run it as long? Have you notice it becoming boiling hot like this?

[simone-silvestri]

Hmmm, not really, I have run only a lower-resolution version of this simulation. I think there are some fixes with CATKE that might have changed something. This looks like a symptom of instability. Does it happen all of a sudden or does the temperature slowly increase?

[taimoorsohail]

It happens throughout the run, starting at around 3 years after the simulation begins. It's in the Caribbean Sea...

[simone-silvestri]

Hmmm, this simulation does not look completely right. The SST should not go above 35-38 degrees even in the first year. Also there seems to be a ton of noise at high latitudes.
I remember there was a problem with GM, what happens if you remove it?

[taimoorsohail]

I will try to remove it and get back to you

[taimoorsohail]

The closures at the moment are:

@info "Defining closures"

using Oceananigans.TurbulenceClosures: IsopycnalSkewSymmetricDiffusivity,
                                       DiffusiveFormulation

eddy_closure = IsopycnalSkewSymmetricDiffusivity(κ_skew=1e3, κ_symmetric=1e3,
                                                 skew_flux_formulation=DiffusiveFormulation())
vertical_mixing = ClimaOcean.OceanSimulations.default_ocean_closure()

closure = (eddy_closure, vertical_mixing)

Do you have recommendations of which to change?

[simone-silvestri]

you should remove the eddy_closure, which is GM + Redi (κ_skew and κ_symmetric, respectively)

[taimoorsohail]

Unfortunately removing GM didn't seem to help - the surface fields are much more noisy and the temperature is a lot warmer at the surface too!

[navidcy]

you should remove the eddy_closure, which is GM + Redi (κ_skew and κ_symmetric, respectively)

This is a 1degree config @simone-silvestri. Without GM I doubt the stratification will be right... Isn't it?

[simone-silvestri]

Indeed, it was just a test to see if there was a specific issue with GM which is the least tested feature in this config. I guess it is not a GM issue... Even if I am not really a fan of it, you should try then to add some viscous dissipation to see if it fixes those maxima

[taimoorsohail]

FYI - I computed surface latent and sensible fluxes and the latent fluxes seem to jump in the first month and stay high - mimicking SST maxima.

I will double check with ERA5 latent fluxes soon.

[simone-silvestri]

Is this integrated or the maximum?