Re: Preliminary advice on finite volume methods

Dear colleague,

There is a paper of B. P. Leonard in which he bypasses the CFL limit
through a different choice of departure stencil in a Lagrangian scheme.
In short, you can use CFL greater than 1, at least in 1D problems,
since you select the apropriate interpolation stencil in the previous
time step.

I can't remember the exact reference now, but I bear the subject very
clearly in memory and I don't have, by now, access to search engines. A
search with 'Leonard', 'CFL', etc, must give you a hint of what it is
exactly.

Another choice is digging in this link
http://www.cfd-online.com/Wiki/Schemes_by_Leonard_-_structured_grids#SHARP_-_Simple_High_Accuracy_Resolution_Program

HTH

Hélio C. Bortolon

Alessandro wrote:
> Well, there are many reasons why I dont want to split river links:
> - one of these is that the hydraulic module of the simulation program uses these links, and I'd rather use the same links for advection (i did not wrote the hydraulic module, but it is robust, stable, tested...)
>
> besides this, there are some other reasons: - the simulation program uses as input a Digital Elevation Model of a watershed and this is made of very easy_life_making square cells - I dont know anything about mesh refining - the time available to work on the code is almost up - etc...
>
> With my big surprise in 1&1/2 days I managed to set up a simple LaxWendroff+VanLeerlimiter in 1D with constant velocity, and, with my big surprise, it is truly flux-conservative! even with low Courant numbers! Almost magic....
> I'll try to insert u(x) in the solution and once I have it working, I'll try to put it into the real code.
>
> Thanks for the help and the suggestions!
> A

.