#807: r.watershed doesnt consider longer distance to diagonal neighbouring pixels
---------------------+------------------------------------------------------
Reporter: aread | Owner: grass-dev@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 mmetz):
Replying to [comment:11 helena]:
> this is where the negative values are:
> http://skagit.meas.ncsu.edu/~helena/grasswork/accum_mfd_neg.png
> [[BR]]
> why does it switch from positive to negative in the middle
> of the slope even for the flow accumulation that apparently originates
> on the top of the hill well inside the area?
The main fow originates from the top of the hill, but cells on the slope
also receive flow from other neighbouring cells with higher elevation.
Tracing this back, particularly for MFD flow dispersion, can lead to the
edge of the region
> I can imagine the high accumulation values starting from outside,
> but the low negative accumulation values (in tens of cells) close to the
center
> of the region have no way of starting outside.
Yes, they have, although the absolute amount of surface flow making it
from the edge of the region to there is minimal, <<< 1. If the edge of the
region has higher elevation than the center of the region, and if there is
an ever so slight downhill slope leading from the edge to the center, some
edge flow will reach the center.
Long explanation:
As the manual says, "Negative numbers indicate that those cells possibly
have surface runoff from outside of the current geographic region." or
from cells with unknown elevation (MASKed or NULL), and this is what the
results should show, if not, it's a bug.
The general slope direction in elev_lid792_1m is from about Northwest to
Southeast and the terrain is quite flat. Many hillslopes seem to be
neither strongly convex nor concave in tangential direction, and in MFD
mode, flow is distributed to many neighbours. Therefore it can be expected
that the sign of flow accumulation will also be spread considerably.
Still, is it true that so many cells receive offmap flow or is it a bug?
In elev_lid.acc.mfd.png
http://skagit.meas.ncsu.edu/~helena/grasswork/accum_mfd_neg.png
only negative accumulation values are displayed, overlayed by
elev_lid792_cont1m (1m contour lines). The negative sign is not propagated
uphill (except for egressing depressions), but, starting from the borders
of the region, slowly descends downhill (edge between coloured and white
cells).
Looking at negative sfd accumulation values, displayed in
elev_lid.acc.sfd.highlight.png
http://sites.google.com/site/markusmetzgiswork/elev_lid.acc.sfd.highlight.png
one of the crucial differences is where the red ellipse is located in the
Northwest. Here, the hillslope seems to be rather flat in the tangential
direction, if not convex, and surface flow including the negative sign is
spread quite far.
Comparing elev_lid792.acc.sfd.zoom.png
http://sites.google.com/site/markusmetzgiswork/elev_lid.acc.sfd.zoom.png
and elev_lid792.acc.mfd.zoom.png
http://sites.google.com/site/markusmetzgiswork/elev_lid.acc.mfd.zoom.png
both show only negative flow, it is apparent that SFD struggles to find a
reasonable flow direction on this hill and shows a straight line where the
negative sign is propagated from the region's edge downhill into the main
basin. MFD on the contrary spreads out the surface flow, typical for
convex hills and similar to flat areas. The negative sign is spread
together with the flow, and because this is at the top of the main basin
in the current region, the negative sign is propagated wide and far
downhill.
Negative flow also appears within the region midway down the large hill in
the Northwestern quarter, keeping in mind that this is not really a hill.
Elevation difference from the top of the hill to the first occurence of
negative flow is about 60cm and from that point the elevation difference
slowly increases. AFAICT, the negative sign is only propagated downhill.
If you are confident that there are indeed negative flow accumulation
values where there should be positive flow accumulation values, please
post the exact coordinates, so I can look in more detail what is going
there.
Markus M
--
Ticket URL: <https://trac.osgeo.org/grass/ticket/807#comment:14>
GRASS GIS <http://grass.osgeo.org>
