[GRASS-user] r.terraflow -> streams ?

[Dylan Beaudette]

On Thursday 18 May 2006 01:07, orkun wrote:
> THANK YOU
>
> Dylan Beaudette wrote:
> >On Wednesday 17 May 2006 00:42, orkun wrote:
> >>Dylan Beaudette wrote:
> >>>On Tuesday 16 May 2006 02:59, orkun wrote:
> >>>>hello
> >>>>
> >>>>can I  produce stream map using r.terraflow maps ?
> >>>>
> >>>>regards
> >>>
> >>>here is one such example:
> >>>
> >>>http://casoilresource.lawr.ucdavis.edu/drupal/node/166
> >>>
> >>>cheers,
> >>
> >>thank you
> >>
> >>I looked the site you directed me.
> >>Compared to r.watershed, r.terraflow produced stream map much faster. But
> >>there  seems to be broken streams. Although I lowered the accumulation
> >>threshold value, broken streams still exist.
> >>How can I get unbroken streams using r.terraflow.
> >
> >Hi,
> >
> >You could try computing the line lengths, and throwing out features that
> > are less that a given length:
> >
> >v.to.db ...
> >v.extract ...
>
> In fact, I want to join broken lines to continuous lines, which are from
> head of
> upstream to downstream.

Hmm.. that might be a little bit harder. It might be simpler to extract _more_ 
lines than required, from the output of r.terraflow, and then remove the 
extra lines that you do not want.

one possible way with with some raster commands:

r.terraflow elev=elev direction=dir filled=filled tci=tci acc=acc 
swatershed=watershed
#get the watersheds
r.to.vect in=watershed out=watershed feature=area -s -v
#
#two possible ways to extract a very rough stream network
r.mapcalc "streams = if(tci >= 10, 1, null())"

#once you have your streams, then experiment with r.grow / r.thin to get the 
best looking stream network.


> >Or.. it would be even better to characterize stream networks according to
> >their "order" , based on the number of contributaries feed into the main
> >channel. Then, keep all streams of order _n_ or greater.
> >
> >>Since I worked on DEM quality, I saw some interesting features
> >>in the other part of the site. Does "DEM vs. Field Slope" show
> >>DEM's quality.
> >
> >I think that you might be referring to this image:
> >http://casoilresource.lawr.ucdavis.edu/drupal/node/189?size=_original
> >
> >The title is a little confusing, as it is a comparison between slope
> > values measured in the field with a clinometer compared with slope values
> > computed from a DEM via r.slope.aspect . The take-home message from that
> > graph was that the DEM-computed slope values tended to under estimate
> > field measured values [1]. Also that a comparison between DEM and field
> > slope seemed to work best when comparing average slope within a 20 meter
> > radius as compared to using the DEM derived slope value at the exact
> > location of the field measurement.
>
> simply ground truth ? if so, what a coincidence !
> All I want to do is to choose  terrain models that are
> produced with different  parameters and  methods
> reflect real ground conditions.
> that is to say:
> which interpolation method for contour elevation data?
> and which grid spacing ?

that is the topic of _many_ articles in the literature. I would suggest the 
book Terrain Analysis for a good background:

Hutchinson, M.F. & Gallant, J.C. Wilson, J.P. & Gallant, J.C. (ed.) Digital 
elevation models and representation of terrain shape Terrain Analysis, John 
Wiley and Sons Inc., 2000 .

> >[1] note that field measurement of slope is not an exact science, and is
> >certainly not conducted at the same scale at which the DEM for this area
> > was created (10 meter grid) .
>
> Is "10 m grid" your preference to create slope map. Actually it
> is my preference. As far as I see coarser grid than 10 m cannot
> some morphological features such as slope, curvature, streams
> and etc.
>
> If you have time, I want to know your ideas.
>

All I had to work with was a 10 meter grid... you might try r.resamp.rst for a 
finer grid ... however your results may vary.


> I hope, grass will have krig interpolation feature
> in next version.

you should be able to do this via the GRASS->R interface described in the 
GRASS Newsletter vol. 3 (http://grass.itc.it/newsletter/GRASSNews_vol3.pdf)

cheers,

-- 
Dylan Beaudette
Soils and Biogeochemistry Graduate Group
University of California at Davis
530.754.7341