Why not use the actual freezing temperature for MinimumTemperatureSeaIce?

[glwagner]

The main benefit is simply that it makes sense rather than being random (and assuming S = 35) or something.

ClimaSeaIce has a simple tool to diagnose the melting temperature so its not hard at all.

ClimaOcean.jl/src/OceanSeaIceModels/minimum_temperature_sea_ice.jl

Line 20 in b3b44b7

MinimumTemperatureSeaIce() = MinimumTemperatureSeaIce(-1.8)

cc @simone-silvestri wondering if there is a reason for this or just convenience?

[glwagner]

side node is the function has the wrong argument order for a mutating function that uses dispatch (should be mutated_args..., dispatch_args... other_args...).

ClimaOcean.jl/src/OceanSeaIceModels/minimum_temperature_sea_ice.jl

Lines 22 to 26 in b3b44b7

	function limit_fluxes_over_sea_ice!(grid, kernel_parameters, sea_ice::MinimumTemperatureSeaIce,
	centered_velocity_fluxes,
	net_tracer_fluxes,
	ocean_temperature,
	ocean_salinity)

[glwagner]

Shouldn't we also prevent the ocean from going below the freezing temperature btw? Not just limit fluxes?

[simone-silvestri]

Nono, it is just convenience. I am not sure how "physical" this method is, but we can use a freezing temperature. If we use a correct freezing temperature indeed, we need to limit the interior. Using a constant temperature we should not need it as if the model is conservative (which is not because of the linear free surface) it should not be able to go below that minimum temperature in the interior if we limit the fluxes.

[glwagner]

The point is not to make it more physical but just make it easier to understand. -1.8 is random whereas the freezing temperaure, is, well, the freezing temperature... it's more logical...

[glwagner]

The model can go below freezing transiently, how would it warm back up? The dynamics here are identical to if we had a sea ice model, cooling causes the temperature to dip below freezing. Sure, you can shut off fluxes after that, but you'd also have to add warming to return the temperature to above freezing.

[simone-silvestri]

Right here, the warming is allowed. It is probably not extremely physical because the warming should be subject to a different albedo if we want to go the "physical" route.

[glwagner]

Grid cells below the surface can dip below freezing due to vertical mixing

[glwagner]

We should avoid assumptions based on how we think things will happen. It's helpful to consider extreme cases. The model has to produce reasonable behavior all the time. It's not about what 'we' want to do but ,along this option as reasonable as possible for anyone

[simone-silvestri]

Ok, let's add the freezing temperature from ClimaSeaIce.

[glwagner]

Ok!

It occurs to me that one may not want to auto-adjust internal ocean temperatures to freezing because the model will no longer conserve energy. Just turning off fluxes maintains conservation. So if we want to adjust ocean temperatures to freezing, it should probably be an option rather than the only possibility (and maybe we don't need it at all).

As an example. I ran a simulation that accidentally removed continents, and the ocean temperature plunged to -40...

[simone-silvestri]

Yeah, longwave radiation is quite powerful! 😄

[glwagner]

Also the atmospheric temperature over siberia is pretty cold.