[GRASS-user] r.watershed: output basin map

GRASS GRASS user
1

   The attached pdf shows the basin map output by r.watershed based on a 10m
DEM and a threshold of 150000. The NHD level 6 subbasins are outlined in
black and the 1:24K NHD streams are in blue.

   The small white areas at the bottom left and center right appear to be
external to defined sub-basins. But I don't know how to interpret the larger
white areas with streams in them (center and top). The top margin is an
international boundary so I can understand not being able to completely
define sub-basins across that boundary since the DEM doesn't reach that far
north, but it's not consistently white.

   Please help me interpret the white areas with vector streams.

TIA,

Rich

(attachments)

ebeco-basins.pdf (31.7 KB)

2

The attached pdf shows the basin map output by r.watershed based on a 10m
DEM and a threshold of 150000. The NHD level 6 subbasins are outlined in
black and the 1:24K NHD streams are in blue.

The small white areas at the bottom left and center right appear to be
external to defined sub-basins. But I don’t know how to interpret the larger
white areas with streams in them (center and top). The top margin is an
international boundary so I can understand not being able to completely
define sub-basins across that boundary since the DEM doesn’t reach that far
north, but it’s not consistently white.

Using the 10m DEM (I suppose your regions settings are the same?) then each pixel is 100 sq.m. So a threshold of 150000 would delineate watersheds of a minimum 15,000,000 sqm = 15 sq km. Does that seem to be what you’re getting? If so, then white areas are small catchments that are hydrologically separated from the larger basins, and thus don’t “belong” to any of the NHD larger basins. The streams in those smaller catchments flow off the map into some other large basin that doesn’t appear in the NHD group.

I’m not sure if this helps any. What do you see as a problem?

··· 
-- 
Micha Silver
052-3665918
3

Hi,

On Thu, Apr 19, 2012 at 12:48 AM, Rich Shepard <rshepard@appl-ecosys.com> wrote:

The attached pdf shows the basin map output by r.watershed based on a 10m
DEM and a threshold of 150000. The NHD level 6 subbasins are outlined in
black and the 1:24K NHD streams are in blue.

The small white areas at the bottom left and center right appear to be
external to defined sub-basins. But I don’t know how to interpret the larger
white areas with streams in them (center and top). The top margin is an
international boundary so I can understand not being able to completely
define sub-basins across that boundary since the DEM doesn’t reach that far
north, but it’s not consistently white.

Please help me interpret the white areas with vector streams.

Please post information about the region, r.info of the map and the command line of r.watershed that you use. Also, do you use a mask?
Pankaj Kr Sharma, please avoid to reply attaching the whole digest.

Thanks,
madi

–
Ing. Margherita Di Leo, Ph.D.

4

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

Please post information about the region, r.info of the map and the
command line of r.watershed that you use. Also, do you use a mask? Pankaj
Kr Sharma, please avoid to reply attaching the whole digest.

madi, Micha, Pankaj:

   The Level 6 HUC basins (a.k.a. HUC-12 or sub-watersheds) are defined by
the USGS to enclose areas of 10k-40k acres (15.625-62.5 square miles;
40.5-161.9 square kilometers). These are a bit too large, but vector
boundaries for 14- or 16-digit hydrologic units are not available for this
area (or for my other project area which is approximately 125 square miles,
323.75 square kilometers) so I tried different threshold values for
r.watershed and 150000 seems to provide additional resolution without
generating a separate drainage for each individual stream segment. Yes, it's
arbitrary. But for the purposes of these two projects it is a practical
balance between too coarse and too fine.

   The region was set by the base, 10m DEM:

projection: 99 (Lambert Conformal Conic; units in International Feet)
zone: 0
datum: nad83
ellipsoid: grs80
north: 735780
south: 655650
west: 2042100
east: 2138580
nsres: 30
ewres: 30
rows: 2671
cols: 3216
cells: 8589936

   The r.watershed command line (all on one line but wrapped in the MUA):

r.watershed -f elev=dem10m.avg acc=dem.acc drain=dem.drain bas=dem.basin
stream=dem_str.seg thresh=150000 --o

   And r.info for dem.basin:

    Type of Map: raster Number of Categories: 94
    Data Type: CELL
    Rows: 2671
    Columns: 3216
    Total Cells: 8589936
         Projection: Lambert Conformal Conic
             N: 735780 S: 655650 Res: 30
             E: 2138580 W: 2042100 Res: 30
    Range of data: min = 2 max = 94
    Data Source:
     dem10m.avg
    Data Description:
     generated by r.watershed
    Comments:
     Processing mode: MFD
     Memory mode: All in RAM
     r.watershed -f elevation="dem10m.avg" accumulation="dem.acc" drainag\
     e="dem.drain" basin="dem.basin" stream="dem_str.seg" threshold=15000\
     0 convergence=5 memory=300

   If the white areas with vector streams (the blue lines) represent basins,
then that's OK. What puzzles me is why a couple of basins would be white
while the others are different colors and the small white areas without
mapped streams are appently external to the basins of interest.

   As far as I know, there's no attribute table for raster maps that hold
descriptions of the generated basins, only information on the contents of
each cell.

Rich

5

On Thu, Apr 19, 2012 at 5:58 PM, Rich Shepard <rshepard@appl-ecosys.com> wrote:

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

Please post information about the region, r.info of the map and the
command line of r.watershed that you use. Also, do you use a mask? Pankaj
Kr Sharma, please avoid to reply attaching the whole digest.

madi, Micha, Pankaj:

The Level 6 HUC basins (a.k.a. HUC-12 or sub-watersheds) are defined by
the USGS to enclose areas of 10k-40k acres (15.625-62.5 square miles;
40.5-161.9 square kilometers). These are a bit too large, but vector
boundaries for 14- or 16-digit hydrologic units are not available for this
area (or for my other project area which is approximately 125 square miles,
323.75 square kilometers) so I tried different threshold values for
r.watershed and 150000 seems to provide additional resolution without
generating a separate drainage for each individual stream segment. Yes, it’s
arbitrary. But for the purposes of these two projects it is a practical
balance between too coarse and too fine.

The region was set by the base, 10m DEM:

projection: 99 (Lambert Conformal Conic; units in International Feet)
zone: 0
datum: nad83
ellipsoid: grs80
north: 735780
south: 655650
west: 2042100
east: 2138580
nsres: 30
ewres: 30
rows: 2671
cols: 3216
cells: 8589936

The r.watershed command line (all on one line but wrapped in the MUA):

r.watershed -f elev=dem10m.avg acc=dem.acc drain=dem.drain bas=dem.basin
stream=dem_str.seg thresh=150000 --o

And r.info for dem.basin:

Type of Map: raster Number of Categories: 94
Data Type: CELL
Rows: 2671
Columns: 3216
Total Cells: 8589936
Projection: Lambert Conformal Conic
N: 735780 S: 655650 Res: 30
E: 2138580 W: 2042100 Res: 30
Range of data: min = 2 max = 94
Data Source:
dem10m.avg
Data Description:
generated by r.watershed
Comments:
Processing mode: MFD
Memory mode: All in RAM
r.watershed -f elevation=“dem10m.avg” accumulation=“dem.acc” drainag
e=“dem.drain” basin=“dem.basin” stream=“dem_str.seg” threshold=15000
0 convergence=5 memory=300

If the white areas with vector streams (the blue lines) represent basins,
then that’s OK. What puzzles me is why a couple of basins would be white
while the others are different colors and the small white areas without
mapped streams are appently external to the basins of interest.

You can query the white cells and discover that they are null values. This is because r.watershed delineates only the basins that fall entirely into the computational region. You didn’t provide information about any mask, but you probably either set a too small mask or should enlarge the computational region in order to delineate also the other basins. In addiction, consider to provide at least a small buffer along the ridge line in order to allow the algorithm to work properly.

Regards,
madi

As far as I know, there’s no attribute table for raster maps that hold
descriptions of the generated basins, only information on the contents of
each cell.

Rich

grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

–
Ing. Margherita Di Leo, Ph.D.

6

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

You can query the white cells and discover that they are null values. This
is because r.watershed delineates only the basins that fall entirely into
the computational region.

madi,

   I believe that the basin to the west of the large white one on the top
also crosses the international boundary; I'll check.

You didn't provide information about any mask, ...

   No mask set.

... consider to provide at least a small buffer along the ridge line in
order to allow the algorithm to work properly.

   How do I do this?

Thanks,

Rich

7

On Thu, Apr 19, 2012 at 6:26 PM, Rich Shepard <rshepard@appl-ecosys.com> wrote:

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

You can query the white cells and discover that they are null values. This
is because r.watershed delineates only the basins that fall entirely into
the computational region.

madi,

I believe that the basin to the west of the large white one on the top
also crosses the international boundary;

Actually, I fear that nature doesn’t really care about politic boundaries :wink:

I’ll check.

You didn’t provide information about any mask, …

No mask set.

… consider to provide at least a small buffer along the ridge line in

order to allow the algorithm to work properly.

How do I do this?

Just enlarge the computational region sufficiently over the expected boundary of the (sub)basin.

madi

Thanks,

Rich

grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

–
Ing. Margherita Di Leo, Ph.D.

8

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

Actually, I fear that nature doesn't really care about politic boundaries

   No, but national data sets do.

   The basin to the west of the top-most white area also does not include the
mouth of the main stem shown, but it's colored as a basin. Further to the
southwest, is another white, non-basin that should be included.

   Perhaps I need to play more with the r.watershed threshold value.

Thanks,

Rich

9

On Thu, Apr 19, 2012 at 6:44 PM, Rich Shepard <rshepard@appl-ecosys.com> wrote:

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

Actually, I fear that nature doesn’t really care about politic boundaries

No, but national data sets do.

The basin to the west of the top-most white area also does not include the
mouth of the main stem shown, but it’s colored as a basin. Further to the
southwest, is another white, non-basin that should be included.

If I understand, the blue lines are the “natural” rivers, which may differ from the calculated stream network. Maybe could be useful to display the calculated stream network in order to understand the logic that the algorithm may have followed creating those basins.
Further, I would also give a try using r.stream.basins.

madi

Perhaps I need to play more with the r.watershed threshold value.

Thanks,

Rich

grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

–
Ing. Margherita Di Leo, Ph.D.

10

On Thu, 19 Apr 2012, Margherita Di Leo wrote:

If I understand, the blue lines are the "natural" rivers, which may differ
from the calculated stream network. Maybe could be useful to display the
calculated stream network in order to understand the logic that the
algorithm may have followed creating those basins. Further, I would also
give a try using r.stream.basins.

madi, et al.:

   I've learned something very useful in pursuing this thread; perhaps it
will be of value to others, too.

   Even though I ran 'r.mask -r' to remove the mask on the DEM, it persisted
and caused r.stream.extract to choke and die because the accumulation map's
region did not match the DEM's region. The reason turned out to be the
unintentional mask.

   So, I started over with the overall DEM at 0.305m resolution. With 40
gigacells it was too large to process on this workstation with "only" 4G
RAM. So was a change to 5m resolution. But, 10m resolution works. (Later
I'll re-run the analyses on my laptop with 8G RAM.)

   My analytical areas are the four 12-digit hydrologic unit sub-watersheds
outlined in black on the attached map. So I changed the region:

g.region vect=basins align=dem_proj

then ran r.watershed with the same 150,000 cell threshold as before.
However, the output is quite different now as you can see on the map. The
resulting basins are too numerous to make broad generalizations about
differences; I think that I need to greatly increase the threshold value.

   I've no idea how the DEM retained the mask outline after I removed the
MASK file, but it certainly did and that flawed the model results.

Thanks all!

Rich

(attachments)

ebeco.pdf (82.4 KB)