#807: r.watershed doesnt consider longer distance to diagonal neighbouring pixels
---------------------+------------------------------------------------------
Reporter: aread | Owner: grass-dev at lists.osgeo.org
Type: defect | Status: new
Priority: major | Milestone: 6.4.0
Component: Raster | Version: 6.4.0 RCs
Resolution: | Keywords: r.watershed
Platform: All | Cpu: All
---------------------+------------------------------------------------------
Comment (by helena):
Markus was right - the difference was due to handling elevations as int
versus fp,
so now I got the version with the diagonal fix
and here is the comparison of spatial patterns
grass64RC05 - SFD with integer elevation
range ~ -200000 - +200000
note zig-zag main streams, missing flow accumulation along the road on the
west ridge
http://skagit.meas.ncsu.edu/~helena/grasswork/accum5K_gr64rc5_3d.jpg
http://skagit.meas.ncsu.edu/~helena/grasswork/accum5K_gr64rc5.png
grass65 compiled in Sep 2009, SFD with FP before diagonal fix
range ~ -200000 - +55000 (why so much lower than grass64?)
no zig-zag on main streams, more realistic pattern on streams, lots of
diagonals on hillslopes
http://skagit.meas.ncsu.edu/~helena/grasswork/accum5K_gr65_sep09.jpg
http://skagit.meas.ncsu.edu/~helena/grasswork/accum_5Ksep09.jpg
grass65 compiled Feb 2010, SFD with fp after diagonal fix
range only slightly different, quite different pattern on hillslopes -
note particularly NW section of the watershed where the previously
diagonal flow changed to horizontal
http://skagit.meas.ncsu.edu/~helena/grasswork/accum5K_gr65_feb10c_i.jpg
http://skagit.meas.ncsu.edu/~helena/grasswork/accum_5Kdiag2010.jpg
I can see that somebody might have liked the diagonal biased version
better than the correct one.
grass65 MFD - no difference between sep 2009 and feb 2010
range ~ -200000 - +16000 (lower than SFD as it should be, but still why
such a diff between 64 and 65?)
most realistic overall
http://skagit.meas.ncsu.edu/~helena/grasswork/accum5K_gr65_mfd.jpg
http://skagit.meas.ncsu.edu/~helena/grasswork/accum_5Kmfdi.jpg
Compared to r.watershed in GRASS65, the GRASS64 results look really bad
for this
high resolution data - I assume for lower resolution data the difference
won't be
as stark but still grass64 will be much worse than the grass65 version.
Should grass65 version of r.watershed be backported to grass64?
although the difference in values needs to be explained (it may be mistake
on my side)
and Markus M may have some additional issues that need to be addressed,
what do others think? who makes the decision?
Helena
--
Ticket URL: <https://trac.osgeo.org/grass/ticket/807#comment:9>
GRASS GIS <http://grass.osgeo.org>