Before I forget, I'd like to write down some of the tweaks that I think we need for the idealized LES setups:
We need to decrease the forcing slightly. Almost all the cases for every suite finish with boundary layers that are just a little too deep. So we should probably decrease forcing for most cases so we can trust results all the way to the end.
I think we should add a case (or two?) that lies "between" free convection and "weak wind strong cooling", like "super weak wind, super strong cooling" :-D Basically the combination of convection and shear appears to be both a particularly challenging situation for closures, and it's the most important ocean case. To put numbers for guidance, the final surface TKE in the four day suite is approximately:
Case
Surface TKE after 4 days (cm^2/s^2)
Free convection
2
Weak wind strong cooling
16
Strong wind weak cooling
26
Strong wind
40
So, we could use a case that has, say, 8 cm^2 / s^2 surface TKE after 4 days, and perhaps a case with ~30-35 cm^2 / s^2. Consider that parameterizations like CATKE expresses diffusivities at the product between sqrt(TKE) and a mixing length. The strong variation in TKE means that we also need variation in mixing length to describe these cases.
We may want to generate a 3-day, 5-day, 7-day, and 8-day suite. This way we can train on 3 suites (eg 4, 5, 6 day) and test on two suites on either side (2, 3, 7, 8). More food for thought. This is less crucial.
Before I forget, I'd like to write down some of the tweaks that I think we need for the idealized LES setups:
We need to decrease the forcing slightly. Almost all the cases for every suite finish with boundary layers that are just a little too deep. So we should probably decrease forcing for most cases so we can trust results all the way to the end.
I think we should add a case (or two?) that lies "between" free convection and "weak wind strong cooling", like "super weak wind, super strong cooling" :-D Basically the combination of convection and shear appears to be both a particularly challenging situation for closures, and it's the most important ocean case. To put numbers for guidance, the final surface TKE in the four day suite is approximately:
So, we could use a case that has, say, 8 cm^2 / s^2 surface TKE after 4 days, and perhaps a case with ~30-35 cm^2 / s^2. Consider that parameterizations like CATKE expresses diffusivities at the product between
sqrt(TKE)and a mixing length. The strong variation in TKE means that we also need variation in mixing length to describe these cases.