Hi everyone,
I’ve found a grid-scale oscillation of the vertical diffusivity (Avt = kᵥ) and the stratification, due to the tidal mixing parameterization specific to the Indonesian Throughflow, from Koch-Larrouy et al (2007). This scheme sets the vertical structure of kᵥ to be proportional to N in locations where dN/dz < 0, or to N² where dN/dz > 0.
The cause: suppose we start with N nearly constant with depth. Then dN/dz changes signs often, leading to different kᵥ parameterizations in neighbouring grid cells, one of which produces large kᵥ and the other small kᵥ. The large kᵥ tends to destratify, while high stratification is maintained where kᵥ is low. This causes dN/dz to alternate even more strongly in sign, creating a feedback.
Here’s an example cast:
Has anyone encountered this before? Any suggestions on what to do about it?
My instinct is to use a modified version where I just make kᵥ proportional to N², not to N. I’ve tested this and it does eliminate the grid-scale oscillation.
Side point: I think the code determines the form of kᵥ at a W point based on dN/dz from the T point below the local kᵥ’s W point. Some method to estimate dN/dz at the same point as kᵥ would seem needed.
Thanks,
Geoff