Compute catchments after resolving closed depressions

[benbovy]

Currently catchment numbers are computed within the find_stack subroutine, which is called before LocalMinima in flow routing routines.

I think that in most cases we want catchment boundaries to rather correspond to the flow route paths obtained after resolving closed depressions.

[benbovy]

Also, having integers (e.g., corresponding to the outlet node index) rather than random float numbers for the catchment ids would make more sense (and yield more stable results).

[benbovy]

One benefit of using random numbers might be that one may easily distinguish adjacent catchments even when using continuous color maps, but arguably this should be addressed by using random, discrete color maps.

[LeoZijerveld]

Thanks for considering this. I would think that the benefits of using a meaningful number (i.e. corresponding to the outlet node) outweighs the plotting benefit you describe in your second comment.