[GRASS-user] Problem with half basins becoming fragmented using r.watershed

GRASS GRASS user
1

Hello,

I am new to the list; please let me know if my question should be formatted differently.

I have been running r.watershed on a large (>400 km2) basin using a 10 m resolution dem. I am finding that it produces some half basins that are either fragmented or very small (i.e., just a few pixels). This seems to happen mostly in areas bordering the stream, though I do sometimes get half basins that are also not connected to the stream. I’m attaching a zoomed in screen shot that illustrates some very strange sliver and fragmented basins that are bordering larger, more appropriately delineated half basins. In this case, the inappropriate basins all border the stream.

I have tried running r.watershed using GRASS6.4 and GRASS7. I have also tried using SFD, MFD with a range of convergence values, and the -4 flag. These options produced different half basin delineations, but the problem with sliver basins and basin fragments has not gone away.

I also checked the dem values for the inappropriate half basins, and they are very similar to the values for neighboring normal basins. So at this point, I’m not sure how to fix the problem. Any advice from the GRASS community would be greatly appreciated.

Thank you,

Erin

(attachments)

Screen Shot 2018-07-13 at 9.55.25 AM.png

2

Not sure what’s going on since I’ve not seen this before. But have you tried different threshold values?

On Fri, Jul 13, 2018 at 2:06 PM Erin Hanan <ejhanan@gmail.com> wrote:

Hello,

I am new to the list; please let me know if my question should be formatted differently.

I have been running r.watershed on a large (>400 km2) basin using a 10 m resolution dem. I am finding that it produces some half basins that are either fragmented or very small (i.e., just a few pixels). This seems to happen mostly in areas bordering the stream, though I do sometimes get half basins that are also not connected to the stream. I’m attaching a zoomed in screen shot that illustrates some very strange sliver and fragmented basins that are bordering larger, more appropriately delineated half basins. In this case, the inappropriate basins all border the stream.

I have tried running r.watershed using GRASS6.4 and GRASS7. I have also tried using SFD, MFD with a range of convergence values, and the -4 flag. These options produced different half basin delineations, but the problem with sliver basins and basin fragments has not gone away.

I also checked the dem values for the inappropriate half basins, and they are very similar to the values for neighboring normal basins. So at this point, I’m not sure how to fix the problem. Any advice from the GRASS community would be greatly appreciated.

Thank you,

Erin

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

3

Hi Erin,

in my experience the small half-basins are not in error, they
appearance depends only on the particular digital elevation
model you are using. And on the threshold you are using in
r.watershed, of course. Did you tray scanning a wide range
of thedholds to make sense of your results?

A similar issue was discussed in details in these papers of ours,
along with possible solutions to the problem of small basins.
We actually discuss slope units, which, essentially, are half basins
obtained from r.watershed with different theshold values
in different regions of a study area:

https://doi.org/10.5194/gmd-9-3975-2016
http://2018.geomorphometry.org/Alvioli_Marchesini_Guzzetti_2018_geomorphometry.pdf

the method presented in the papers is based on the software
r.slopeunits, available here:

http://geomorphology.irpi.cnr.it/tools/slope-units

and one can get simple half basins by ignoring the option "cleansize",
or specifying from value for "cleansize", which
would remove small slope units form the final map using a few different methods.

hope this helps.

Massi

2018-07-13 19:06 GMT+02:00 Erin Hanan <ejhanan@gmail.com>:

Hello,

I am new to the list; please let me know if my question should be formatted
differently.

I have been running r.watershed on a large (>400 km2) basin using a 10 m
resolution dem. I am finding that it produces some half basins that are
either fragmented or very small (i.e., just a few pixels). This seems to
happen mostly in areas bordering the stream, though I do sometimes get half
basins that are also not connected to the stream. I'm attaching a zoomed in
screen shot that illustrates some very strange sliver and fragmented basins
that are bordering larger, more appropriately delineated half basins. In
this case, the inappropriate basins all border the stream.

I have tried running r.watershed using GRASS6.4 and GRASS7. I have also
tried using SFD, MFD with a range of convergence values, and the -4 flag.
These options produced different half basin delineations, but the problem
with sliver basins and basin fragments has not gone away.

I also checked the dem values for the inappropriate half basins, and they
are very similar to the values for neighboring normal basins. So at this
point, I'm not sure how to fix the problem. Any advice from the GRASS
community would be greatly appreciated.

Thank you,

Erin

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

4

2018-07-16 18:53 GMT+02:00 Erin Hanan <ejhanan@gmail.com>:

Hi Erin,

let me forward your email to the list, since this may be interesting to others.

Yes, this helps a lot; thank you! I've been reading through your paper and
see that you mentioned one possible solution would be to remove slope units
smaller than "cleansize", and then enlarge the adjacent slope units to fill
the missing area using the r.grow. I'm curious whether I can also achieve
this using r.area to remove the small half basins and then r.grow. I've also
installed your software (r.slopeunits), and am planning to experiment with
the different approaches today.

By the way, I forgot to mention it in my original email, but yes, I tried
several different threshold values. Interestingly, changing the threshold
tended to change the size of larger half basins, but the sliver and
fragmented basins always remained a problem.

Anyway, thanks again for your very helpful email!

You are most welcome. Also, one of our cleaning approaches does
exactly what you suggest:
remove small areas and fill the empty spaces with r.grow.

Let us know if you get meaningful results!

Massi

5

Thank you again Massi. This fixed my problem!

···

On Mon, Jul 16, 2018 at 10:10 PM, Massi Alvioli <nocharge@gmail.com> wrote:

2018-07-16 18:53 GMT+02:00 Erin Hanan <ejhanan@gmail.com>:

Hi Erin,

let me forward your email to the list, since this may be interesting to others.

Yes, this helps a lot; thank you! I’ve been reading through your paper and
see that you mentioned one possible solution would be to remove slope units
smaller than “cleansize”, and then enlarge the adjacent slope units to fill
the missing area using the r.grow. I’m curious whether I can also achieve
this using r.area to remove the small half basins and then r.grow. I’ve also
installed your software (r.slopeunits), and am planning to experiment with
the different approaches today.

By the way, I forgot to mention it in my original email, but yes, I tried
several different threshold values. Interestingly, changing the threshold
tended to change the size of larger half basins, but the sliver and
fragmented basins always remained a problem.

Anyway, thanks again for your very helpful email!

You are most welcome. Also, one of our cleaning approaches does
exactly what you suggest:
remove small areas and fill the empty spaces with r.grow.

Let us know if you get meaningful results!

Massi