Implement Kinematic Viscosity

[ole]

Ted and Rudy to provide formula for KV forcing term. Tests will show eddy formations.

[ole]

Stephen Roberts reckons it'll be a bit involved as one needs to approximate the second order derivative and this will involve going beyond immediate neighbours. Obviously, this will take some effort to get right especially near the boundaries.

[ole]

New idea: Using the existing architecture we could compute the second order derivatives in two iterations by taking the gradients of gradients. In other words store the first order derivatives when computed and then use the same algorithm to compute the second order derivatives. Information beyond immediate neighbours would therefore be implicit.

[ole]

In phone conversation with Stephen Roberts early September 2008 he suggested that we used the second order derivative already implemented in the smoothing algorithm in ANUGA. In other words we could use the Finite-Element solver to approximate the viscosity term for use with the Finite-volume scheme.

Moreover, the majority of time spent in the smoothing algorithm has to do with the least squares process, so the second order derivative on its own is likely to be rather fast.