Climatemachine.jl: Bomex update

[trontrytel]

@tapios , @akshaysridhar , @szy21 , @charleskawczynski

I have run Bomex in Equilibrium and NonEquilibrium configurations. Both simulations crash after an hour with ERROR: LoadError: abs(δρ) > 0.0001. Not sure if this is to be expected for now? The default setup was to run it for 30 minutes.

Here are results from the first hour:

At 69 minutes both runs have similar cloud top heights and q_liq. At the earlier times the NonEquilibrium model has larger vertical velocities because of larger latent heat release. This is caused by the initial condition differences between the two models, I think.

[charleskawczynski]

Is this the LES or single stack?

[trontrytel]

Is this the LES or single stack?

LES

[szy21]

If you would like to run it longer, simply increase the numbers here (@akshaysridhar we should make this clear in the docs):

check_cons = (
        ClimateMachine.ConservationCheck("ρ", "3000steps", FT(0.0001)),
        ClimateMachine.ConservationCheck("ρe", "3000steps", FT(0.0025)),
    )

These are for 30 min simulations, so I would increase them by e.g. a factor of 12 for a 6h simulation.

[trontrytel]

If you would like to run it longer, simply increase the numbers here (@akshaysridhar we should make this clear in the docs):

check_cons = (
        ClimateMachine.ConservationCheck("ρ", "3000steps", FT(0.0001)),
        ClimateMachine.ConservationCheck("ρe", "3000steps", FT(0.0025)),
    )

These are for 30 min simulations, so I would increase them by e.g. a factor of 12 if you run a 6h simulation.

Thanks! I did not realise there was something else to do when extending the time.

Would be good to make it automatic choice, based on simulation time?

[szy21]

Would be good to make it automatic choice, based on simulation time?

Good point. @charleskawczynski @akshaysridhar any thoughts?

[akshaysridhar]

If you would like to run it longer, simply increase the numbers here (@akshaysridhar we should make this clear in the docs):

check_cons = (
        ClimateMachine.ConservationCheck("ρ", "3000steps", FT(0.0001)),
        ClimateMachine.ConservationCheck("ρe", "3000steps", FT(0.0025)),
    )

These are for 30 min simulations, so I would increase them by e.g. a factor of 12 if you run a 6h simulation.

Thanks! I did not realise there was something else to do when extending the time.

Would be good to make it automatic choice, based on simulation time?

Yeah I think now this should all be updated based on a full nightly test time, for the complete timeend = 6 hour simulation

[charleskawczynski]

I kind of like the conservative/concrete tolerances. Maybe we just add a comment, for now, near timeend about the need to increase conservation tolerances due to the accumulation of errors over time?

[trontrytel]

Final update on this. Here is the time evolution for the two options for the whole 6 hours simulation time:

And here is the last hour average for the two options, compared to the plot from the paper:

I think this looks good. Should come back to it as a nightly test, once we have nightly tests.