[GRASS-dev] Testing results of new r.watershed against r.terraflow

[Isaac Ullah]

Here are some additional results of testing between new version of
r.watershed and r.terraflow.

*Time costs:*

Map A (2.8 million cells), r.watershed is 52 seconds faster than r.terraflow

Map B (21 millions cells), r.watershed is 2 minutes 58 seconds faster than
r.terraflow

*Accuracy in r.landscape.evol simulations:*

Maps produced using r.watershed for flowacc show realistic values of erosion
and deposition *with no smoothing* after 10 model cycles(1 cycle = 1 sim
year). The difference between smoothed and unsmoothed maps is at maximum
only .5 meters, and the diffs are dispersed topographically. Image link:
http://www.public.asu.edu/~iullah/files/MEDLAND/smo_nosmo_diff_terraflw.png

Maps produced using r.terraflow for flowacc show spikes as high as +/- 96
meters (with no smoothing) after 10 model cycles. Spikes are concentrated in
the channels, and diff between smoothed and un smoothed results also show
anamolies are concetrated in the channels. Image link:
http://www.public.asu.edu/~iullah/files/MEDLAND/smo_nosmo_diff_rwshd.png

I am currently testing new r.watershed in a longterm model run (100 cycles)
with no smoothing, and will report the results here. In r.lnadscape.evol for
a 2.8 million cell map, the smoothing routine costs 23.5 seconds. Using new
r.watershed with the smoothing turned on takes 1 minute 5.5 seconds. Without
smoothing takes only 42 seconds. Using r.terraflow with smoothing takes 1
min 57 seconds. Using r.terraflow without smoothing takes 1 minute 33
seconds, but produces unusable results. Also, using r.watershed eliminates
the need to run r.fill.dir before each model run--an additional time
savings...

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/20081203/3addc22b/attachment.html