[GRASS-user] r.stream.extract Montgomery

Markus Metz markus.metz.giswork at googlemail.com
Sun Aug 8 08:24:22 EDT 2010


stephen sefick wrote:
> I would like to use the Montgomery method to extract streams (exp~2
> from the paper).  What should the threshold value be?  And in general
> what should the threshold value be?  I am working in the southeastern
> coastal plain (USA), which is characterized by low gradient (if this
> helps).  I am using 1m resolution LIDAR data.  I would like to extract
> the most realistic map of streams that I can.

There is no universally recommended threshold, the threshold value
depends on how much detail you want. Reference stream maps are often
digitized on a certain scale, showing a more or less detailed stream
network. A threshold value of 1 (cell) doesn't make sense because then
streams will be everywhere. As a general rule of thumb, standard
deviation of the absolute flow accumulation values seems to be a
useful value and should in general provide accurate stream locations.
This general rule of thumb considers what detail can be extracted from
a given DEM, and not where real stream heads are located. I guess that
real stream heads can only be extracted from very high res DEMs.

Channel tracing is not that much of a problem, independent of the
threshold used, the extracted streams should match real streams.

The challenge is the proper determination of stream heads (channel
initiation), which can only be done with ancillary data like
precipitation and soil infiltration rate. It also depends on whether
permanent streams should be extracted or permanent plus temporary
(e.g. seasonal) or permanent plus temporary plus ephemeral (existing
only after heavy rainfall).

The -b option in grass 7.0 (beauty flat areas) applies to flat areas
only, and as the name says, beautifies flat areas. Stream locations
might or might not be more accurate with the -b option. The -b option
applies to flat areas in the DEM, i.e. zero slope and zero tangential
and profile curvature. Coastal plains are not flat areas in this sense
because there should still be slight variation in elevation. The flat
areas targeted with the -b option are lakes, broad rivers and sea.
These areas might need to be flattened in the DEM anyway, depending on
the source DEM, in order to get realistic flow accumulation and stream
networks.

HTH a bit,

Markus M


More information about the grass-user mailing list