Re: [Anuga-user] The NaNs issue

[r r. <rra...@gm...>]

Dear Ole,
Now that the issue boiled down to the way "asc" file is converted to
"pts", one can think of setting up a test to reflect it. But one may
not look for failure error in this case.

Since the whole grid is being wrongly shifted / clipped we can at the
maximum expect wrong representation of the elevation at any given
location, like once Joaquim mentioned the coast seemed to be shifted
in his case. On finer grid resolution this may not be apparent. but at
the resolution used by Joaquim (800m), we can clearly see the effect.
But we need reference pts file to compare. one way to do is use
Mirone's "conv2anugapts" utility. I did not check it after very recent
modifications. But I use another route> namely take the "asc" file and
open in mirone and save it as "grd" file. then use gmt command
"grd2xyz in.grd >out.xyz" to get xyz file. use the hack from Okushiri
example to create pts file. This I verified to be free of pixel vs
grid issueby comparing randum grid locations using mirone.

But another method is also possible
in 1st iteration use chess board type of elevation setup on ideal test
bed and derive profiles at say x/2 and y/2 and also derive the asc
file from it. ( I verified that the anuga produced asc file is
correctly representing original grid when I use the grd-xyz-pts route)

in 2nd iteration use this asc file and look at the profiles again and
compare with the reference.
It should clearly show the displaced step elevations.

May be I can try setting up this ideal method

Does it sound reasonable?
Cheers
-rajaraman