No subject

[Charles Ehlschlaeger]

>From: larch at lamar.ColoState.EDU (Merlyn Paulson)
>Sender: grass-lists-owner at max.cecer.army.mil
>Reply-To: grassu-list at max.cecer.army.mil
>Subject: Re:  R.WATERSHED QUESTION.

>R.watershed has been running since 8:15AM 8/22 (today is 8/26) on the
>following dataset (current progress follows):

>The output from r.watershed will be divided into watershed 
>basins.  There are two possible methods of tabulating the information: 
>1) by only including data pertaining to the basin itself, or 2) using 
>data from the basin, and all basins upstream of it.

>Would you like the data organized:
>1) Basin only
>2) Basin and upstream
>3) Both
>OR 0) to cancell program
 
>Your choice: 1

>OK, seg should start running now using the following form:
>/usr3/grass4.0/etc/water/seg el="elevation at PERMANENT" de="lakes at PERMANENT" 
>t=1 ba="basin1" ar="relate1" ac="accum1"

>SECTION 1 beginning: Initiating Variables. 5 sections total
>    segments in row:    167
>segments in columns:    165
>     total segments:    27555
>      open segments:    213

>SECTION 1b (of 5): Determining Offmap Flow. Percent Complete:  
 
>SECTION 2: A * Search.                      Percent complete:    4%

>MY QUESTION IS THIS:

>DOES IT MAKE SENSE THAT IT TAKES APPROXIMATELY 36 HOURS (SUN IPX) TO
>KNOCK OFF EACH PERCENT IN SECTION 2: A + Search?  

>I really need to get this done, so am extremely hesitant to kill the
>job and begin again.

>I read your previous message about the comparison with Spearfish...
>I don't know how big Spearfish is and don't have immediate access to
>it to find out.

>Thanks for your reply.

>Merlyn Paulson, Prof.
>Landscape Sciences & GIS                         Telephone (303)491-7594
>Colorado State University                         Faxphone (303)491-7745
>Fort Collins, Colorado 80523            E-mail larch at lamar.colostate.edu

The size of the basin thresshold is unimportant for the speed at which
SECTION 2 works. But, unless your disk space is huge (I mean HUGE), the
program will crash when it tries to build the watershed information file
at the end of the run. That is because each watershed basin gets about
80 characters of description in the file you indicated with:

ar="relate1" 

Because the minimum size of basin threshold is one cell (t=1), your basin
file is going to be rows * columns * 80 characters long.  The following part 
of your message:

>SECTION 1 beginning: Initiating Variables. 5 sections total
>    segments in row:    167
>segments in columns:    165
>     total segments:    27555
>      open segments:    213

indicates that the size of the region you are running (number of rows times 
number of columns) is huge. If you run the following command:

g.region -p

you will find the number of rows and columns being analized. The time it
takes the segment version of r.watershed to run is linearly dependent on:
rows * columns. If you double the horizontal and vertical resolution, 
you will reduce the time r.watershed to run by a factor of four. Thus, if
you want the program to finish in approx 36 hours, multiply the horizontal
and vertical resolutions by 10.

Another goal would be to run the ram version of r.watershed. That version
(you can choose the ram version on the command line or the interactive 
r.watershed) runs at least 10 times faster than the segment version.

Chuck Ehlschlaeger
chuck at pollux.geog.ucsb.edu