This is version 4 (or communal version) of CASCADE which I suggest we
call CASCADE2000.

To compile:

make lib
make cascade

Check the compilation flags in the Makefile file as well as the plotting
routines you want to use. There is a set of dummy routines for those who do
not wish any runtime graphics.

This should create two libraries in the LIB sub-directory: libxplot.a and
libnn2d.a. These libraries are the "standard" versions of the X11 plotting
routines developed by Jean Braun and the natural neighbour routines
developed by Malcolm Sambridge and Jean Braun. If you want your version of
CASCADE to remain compatile with further development of these libraries
by the authors, do not attempt to modify them; let us know if there is
a bug...

It is hard to really define what is modified in this version because
I went through a large number of items requested from all in the group.

Among them:
- complete modularity: each routine does only one thing and each argument
in the calling list is defined (they are about 30 separate subroutines now)
- modify the convergence criterion in the iterative solver of the diffusion
equation to allow for very large model runs
- modify the fluvial erosion algorithm to use slope, not height difference
to determine where the water is going (this was already modified in an
updated version of version 2, thanks to Philippe)
- modify the algorithm to deal with local minima to make sure that the
water goes out of lakes through the proper sill and does not go uphill
(again thanks to Philippe for pointing out this problem); the new
version is in fact faster than the old one...
- modify the cascade algorithm to make it more performant
- all outputs are in ASCII and stored in a variety of files in a special
directory (different for each run)
- input can be hard-wired in or read from an input file; this is true for
general parameters and nodal geometry, bc, initial topography, etc...
- erosion parameters (3 in the present version, but it is allowed to have
as many as needed) are stored in a single array as well as what I called
the "memory" of the model (ie all nodal values that have to be memorized
from one time step to the next);
- dynamic updating of the mesh is simplified (as suggested by Philippe)
by the introduction of these two super-arrays.
- graphic calls in the main program to help debugging but a set of dummy
graphic routines for those who dont have an X-windows compatible system
- graphic program has been added that uses plain X-windows calls and should
therefore run on any Unix machine
- debugging facility which, when turned on, tells you where the program 
was interrupted
- more documentation of what is happening inside each routine
- list of subroutines called within each subroutine

As you see, this has been a major rewritting exercise and I cannot promise
that this new code is completely error free. Please use it and let me know
AS SOON AS POSSIBLE where you have encountered problems or noticed mistakes.

The main reason for this new version to exist is that it should provide a
basis for all future improvements, new modules, etc... It is quite essential
that you try to keep the "style" of the code intact if you wish
your additions/modifications to be considered by all.

May I suggest that if an error is found in this version or if a major
improvement can be done to it, you pass it on to me as soon as possible and
I will incorporate it (if everyone agrees). Add-ons should be well
documented so that we can easily incorporate them. You have to document
all changes to the existing routines/main program that are required as
well as provide the new code IN A SUBROUTINE/FUNCTION.

Thank you for your feedback,

Jean Braun
Canberra 2000