[GRASS-dev] testing results of r.watershed2 against old r.watershed

[Isaac Ullah]

Hi all, I'd like to report the results of testing I just did with the new
r.watershed2 module from the addons svn. I am using the latest svn source
version of grass6.4, compiled on a dual processor dell with 2 gigs ram, and
running the latest ubuntu 8.10 OS. I used the dataset that I've used for all
my testing and development for the LandDyn addon modules (up on svn addon
server), and with which I have good knowlege of the results of hydrographic
modeling with r.flow, r.terraflow, and old r.watershed. I A/B tested the old
r.watershed module against the new r.watershed2 module wit this same
dataset, and with same options enabled (except of course for "memory" option
which is only available in new r.watershed2, and which I set to 1000). The
input DEM is approximately 1.5 million cells (15meter resolution), and is of
a single large watershed (areas outside watershed boundaries are NULL).

Results: both modules successfully produce quality output maps. In fact,
when the output maps are compared (subtracted with mapcalc) there is
absolutely NO DIFFERENCE between them. All values are the same. The main
difference then between the modules is the speed at which they operate. Old
r.watershed took about 1.5 minutes to finish, while the new r.watershed2
took ABOUT 5 SECONDS!!! This is an amazing increase in speed for exactly the
same result!

I'd like to put my vote in for replacement of original r.watershed by
r.watershed2 in current GRASS6.4 svn, especially because now I'd like to
replace usage of r.terraflow in the LandDyn scripts with r.watershed2. This
would greatly increase the speed, and perhaps the accuracy of those modules
(r.landscape.evol modules).

Two minor suggestions for improvement, however.

1) neither version of r.watershed propagates NULL values. Currently, one
must use a MASK to make sure wierd values aren't calculated for outmaps in
NULL areas of the input map. A flag to ignore NULL areas would make alot of
sense, and perhaps in GRASS7, make the default option be to ignore NULL's
with a flag to reverse.

2) both versions have redundant options for outputing flow accumulation map:
the output options "accumulation" and "visual" output the same map! The only
difference between them is that "accumulation" lacks a color table (on
display, all values are magenta), while "visual" has been assigned some
custom set of color rules, so that it looks nice when displayed. Since no
time is lost for color rule assinment, it does not make sense to allow for
the option to create two otherwise identical maps. IMO, "visual" output
option should be removed, and "accumulation" option should just output a
properly colored map.

Cheers,

-- 

Isaac I Ullah, M.A.

Archaeology PhD Student,
ASU School of Evolution and Social Change

Research Assistant,
Mediterranean Landscape Dynamics Project
***************************************************
isaac.ullah at asu.edu
ullah at archaeologist.com

http://www.public.asu.edu/~iullah
***************************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/grass-dev/attachments/20081114/8877ae8e/attachment.html