Implicit restoring

[jagoosw]

I frequently use restoring on tracers and often find that the restoring can limit my timestep when implemented as a forcing while keeping values positive. To remedy this I recently implemented it in an implicit form that can be passed into Biogeochemistry as a modifier (updates the field during update_biogeochemical_state!) which is unconditionally positive.

I wonder if this would be good to add to the source code here since it seems like a widely useful addition? I guess it could also go in Oceananigans although I'm less sure of the general utility?

@johnryantaylor @glwagner

[glwagner]

I think anything that can be construed as a change to a time-stepping scheme or solver should be implemented in Oceananigans. Generally we have to rely on rule-based thinking for such decisions rather than trying to guess "is there an application?" because we just don't know what all the possible applications might be. Everyone benefits from this strategy in general I think.

In terms of ways to implement such an interface -- there is a bit of complexity associated with the fact that you'd want to do something different depending on whether or not a tridiagonal solver is being used for a closure. I think such things can be figured out though.