>For the distances between cell centres, yes.
>It feels like you're performing some operations that aren't particularly finite element-y?
True. I try to mimic a finite volume (two-point flux approximation) scheme here.
>Would you be using the approximate gradient in a variational form immediately, or are there further >calculations you want to carry out?
I would want to use it directly.
>In the former case, I think there's a nice FE way to get what you want.
I'm not bound to the FV way here. I just want to solve the attached PDE system. Doing it the FE way would have the additional benefit of having a lot more freedom discretisation-wise. Doing it the FV way, I just happen to know what kind of approximation works
(upwinding of the relative permeabilities kr\alpha(Sw), fluxes across element faces)
In the latter case, we can probably hack something up if we're smart enough...