[GRASS-user] Merge spatially connected features

[Markus Metz]

Hi Johannes,

IIUC, what you want to do is an operation that involves topological
relations of vector geometries (connected lines) and a common attribute.
There is no easy common recipe for this.

Just a suggestion:
for each stream order:
  extract all lines with this stream order (v.extract)
  identify connected lines (v.net + v.net.components)
  update a new attribute of the original lines with the comp attribute of
the output of v.net.components plus some offset to separate different
stream orders

HTH,

Markus M


On Tue, Mar 10, 2020 at 5:20 PM Johannes Radinger <
johannesradinger at gmail.com> wrote:
>
> So...no also with GRASS-user as recipient...
>
> On 05.03.20 16:21, Micha Silver wrote:
> >
> > On 3/5/20 10:47 AM, Johannes Radinger wrote:
> >>
> >> Hi Micha, hi all,
> >>
> >> sorry for my late response...however, just today I managed to try
> >> your approach of building polylines to connect "touching stream
> >> lines"...but...
> >>
> >> On 24.02.20 16:48, Micha Silver wrote:
> >>>
> >>> On 24/02/2020 10:45, Johannes Radinger wrote:
> >>>> Hi all,
> >>>> I have a large river network dataset (lines). Now I'd to assign
> >>>> unique categories to each group of connected lines that have an
> >>>> attribute in common.
> >>>>
> >>>> For example, my rivers are categorized based on some kind of stream
> >>>> order. I want to group all rivers that belong to stream order 2 and
> >>>> are spatially connected; each group should get a unique category
> >>>> value. I thought that I could first extract all rivers with a
> >>>> particular attribute (e.g. stream order = 2) which will provide me
> >>>> some scattered pattern of lines. Then I need a spatial join tool to
> >>>> make subgroups of lines that are connected. How can I achieve the
> >>>> latter? Any idea?
> >>>
> >
> >
> >>>
> >>> Here's a procedure that might work for you. Somewhat clunky, but I
> >>> think it gets what you want.
> >>>
> >>> It's based on the v.build.polylines module to connect all touching
> >>> stream reaches. First extract each order from the stream vector into
> >>> a new vector. Then build polylines. Patch them all together. Now you
> >>> have a polyline vector with a single cat value for each set of
> >>> original stream reaches that had the same order and that were
touching.
> >>
> >> Unfortunately, the v.build.polylines tool does not work as it only
> >> does not connect multiple (intersecting) lines like in a river
> >> network. As an example I tried to build polylines from the stream
> >> network of the NC dataset. Yous suggested approach should result that
> >> each sub-network (i.e. river network that is not connected to another
> >> one) should get its own ID/cat...however, v.build.polylines results
> >> in a connected stream network that consists of multiple cats:
> >>
> > Maybe I misunderstood your question. The steps I tried use a
> > stream_order column to group stream segments, then apply a new
> > attribute "merged_id" to those stream orders that touch. i.e. that
> > connect to the same confluence point.
> >
> >
> > Here's what I get using the nc_basic_spm mapset:
> >
> >
> > r.watershed elev=elevation accum=nc_facc drain=nc_fdir bas=nc_bas
> > stream=nc_str thresh=1000
> > r.stream.order stream_rast=nc_str direct=nc_fdir elev=elevation
> > accum=nc_facc stream_vect=nc_streams
> > ORDERS=`v.db.select -c nc_streams group=strahler column=strahler`
> > echo $ORDERS
> >
> > # Create a new stream vector for each stream order
> >
> > for o in $ORDERS; do
> >
> >     v.extract input=nc_streams output=streams_${o} where="strahler=${o}"
> >
> >     # Give each polyline it's own cat value
> >
> >     v.build.polylines input=streams_${o} output=streams_${o}_polyline
> > type=line cat=first
> >
> > done
> >
> >
> > # patch the stream orders back together
> >
> > POLYLINES=`g.list vect pattern="streams*polyline" separator=comma`
> >
> > v.patch input=$POLYLINES output=streams_polylines
> >
> > v.db.addcolumn map=streams column="merged_id INTEGER"
> >
> >
> > # And use v.distance to update that merged_id column from cat values
> > in polylines vector
> > v.distance from=streams to=streams_polylines upload=cat column=merged_id
> > v.db.addcolumn map=nc_streams column="merged_id INTEGER"
> > v.distance from=nc_streams to=streams_polylines upload=cat
> > column=merged_id
> >
> > Now, all stream reaches that have the same order and are "touching"
> > have the same merged_id. See the attached image.
> >
> >
> > If that's not your purpose, then just ignore...
> >
> Micha thank you for your help and of course, you're fully correct!
> Merging lines that belong to the same stream order works in this case
> well...but this is because of the definition of the Strahler ordering
> system, where there is only one "touching node" (i.e. river junction) of
> two rivers of the same stream order (i.e. when two 2nd order streams
> meet, the become a 3rd order stream). Thus your solution works because
> of this specifics and might not work if streams are grouped based on a
> different (ordering) system.
>
> I was already thinking of the next step (beyond simple Strahler): As
> mentioned in my initial post I am dealing with "some kind" of stream
> order. It is similar to grouped stream orders (e.g. stream order 1-2 =
> "headwater streams"). I tried to somehow reproduce my situation based on
> your example of the NC dataset. What I basically did was to reassign a
> new stream order "99" to all former 1st and 2nd order streams. Then I
> did exactly what you did in your example, and of course I don't unique
> merged_ids for the subnetworks of touching lines (see attached Figs)
> that all belong the the same "order" 99 (the original strahler order 3
> works of course, see Fig.)...So is there a more general way (as said
> something like v.dissolve but for lines/networks?):
>
> #####################
> g.region raster=elevation
>
> r.watershed --o elev=elevation accum=nc_facc drain=nc_fdir bas=nc_bas
> stream=nc_str thresh=1000
> r.stream.order stream_rast=nc_str direct=nc_fdir elev=elevation
> accum=nc_facc stream_vect=nc_streams
>
> #ORDERS=`v.db.select -c nc_streams group=strahler column=strahler`
> #echo $ORDERS
>
> # Regroup orders 1-2 (to 99)
> v.db.addcolumn map=nc_streams at test2 columns="strahler_groups INTEGER"
> v.db.update map=nc_streams column=strahler_groups query_column=strahler
> v.db.update map=nc_streams column=strahler_groups value=99
> where="strahler=1 OR strahler=2"
>
> NEWORDERS=`v.db.select -c nc_streams group=strahler_groups
> column=strahler_groups`
> echo $NEWORDERS
>
> # Create a new stream vector for each stream order
> for o in $NEWORDERS; do
>      v.extract input=nc_streams output=streams_${o}
> where="strahler_groups=${o}"
>      # Give each polyline it's own cat value
>      v.build.polylines input=streams_${o} output=streams_${o}_polyline
> type=line cat=first
> done
>
> d.vect -c map=streams_99_polyline at test2
> #################
>
> Thank you very much!
>
> Cheers,
>
> Johannes
>
> >
> >> v.clean --overwrite input=streams at PERMANENT output=streams_break
> >> tool=break
> >> v.build.polylines --overwrite input=streams_break at test
> >> output=streams_poly cats=first type=line
> >> d.vect -c map=streams_poly
> >>
> >> So what would be needed here is some kind of tool that connects all
> >> touching lines and assigns a common category value, similar to the
> >> v.dissolve tool for polygon features. I can imagine that such a task
> >> might be not that uncommon also in another context? Any suggestions
> >> how to achieve this in GRASS?
> >>
> >> A workaround that came into my mind was to create buffers around
> >> lines in order to make areas out of lines. Subsequently these
> >> touching areas can be merged using v.dissolve and the information
> >> about the common category can be queried using v.distance.
> >> Nevertheless, a rather cumbersome way to just assign a common
> >> category value to all lines that are touching...
> >>
> >> Any further ideas?
> >>
> >> cheers,
> >>
> >> Johannes
> >>
> >>>
> >>>> Cheers,
> >>>> Johannes
> >>>>
> >>>> _______________________________________________
> >>>> grass-user mailing list
> >>>> grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org>
> >>>> https://lists.osgeo.org/mailman/listinfo/grass-user
> >>>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-user
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20200311/6091035d/attachment.html>