r.watershed questions from B.Haney

[MORGAN at GAMMA.IS.TCU.EDU]

									TCU Ft.Worth, Tx.
									12/8/93
Howdy interested readers,
		I am a Grad student at TCU in Ft. Worth, Tx.  My thesis work includes using the r.watershed program of GRASS.  I would appreciate your helpe generating better slope length for use in a RUSLE GIS.  My local sources of GRASS expertise at the COE Ft. Wort
h District office and SCS Carto/GIS center have had little experience with r.watershed, and in generating the slope length and slope steepness.  My class and thesis experience on GRASS have not, until this point, been with r.watershed.  Would you help by
 taking a few minutes to answer these questions?
	Let me describe my environment and data sources to give you some background-
I am running 4.0 on a SUN 4/110 with 8 megs of memory, and a 1.2gig hard drive, and will use r.mapcalc to multiply the RUSLE factors/layers together.  I am using TM data for the landuse factor, and SCS soil survey data for the erosion factor on a 100,000
 acre watershed with little relief --  at a few points the elevation changes from ~800 to 1050 ft, but generally the elevation varies from 800 to 550 ft over a few miles.  My elevation layer was resampled from USGS DEM (100 m resolution) elevation data w
ith 1 meter elevation values relative to the West Dallas area to cell size @ 28.5 m resolution N-S and E-W.  After contouring with r.surf.contour, I attempt to run r.watershed on each of two halves (N and S).  The results show about 75% with a slope leng
th of 0.03 (category 3) and most of the remainder spread between 0.025 and 1.350 (slopes from <1 to 10 degrees) except for a few points where the values run from 2.00 to 5.00.
	I have little experience with determining slope and the ls factors so I checked with my main professor, Dr. Ken Morgan.  He has experience with aerial photo and satellite slope interpretation, and thinks I should be able to get better results.  My three
 questions are: 
	1. What  is the best way to represent disturbed land, and is it required since the landuse factors are covered in another portion of the equation?  I have been using a value of 10% for every cell (my local SCS representative believes it's accurate).  I 
could generate a new layer for this input from my CP (landuse) layer, but would this accentuate the slope at the small areas of truely disturbed land?
	2. How important is the depression map, and would recognized bodies of water (streams, lake and ponds) from my landuse raster map provide enough data to influence the results?
	3.   The main one - For max slope length the resulting ls layer closely follows the elevation's contour lines and shows a zero ls value between the lines, whereas I would rather have a general slope over a whole area.  How would you suggest modifying th
e data, or input values?  I use a slope length of 100m but the changes between contours happen in one cell rather than showing a change in slope over a distance.
	When I run r.watershed, I give input names for basin, relationships, accumulation, with a 100,000 acre exterior basin value.  I do not give input for a depression map, blocking map, flow, etc. or other lumped hydrological modeling, and I ask for basin o
nly on the LS and S factors.  The program runs for days without finishing on the whole watershed so I split it into a northern section and a southern section that run on my machine for about 1.5 hours before a segmentation fault occurs.  I am not worried
 about the segmentation fault because my LS and S resultant maps have been generated then and they are the only results in which I am interested.
	GRASS 4.1 and 3.0 are available to me, but I would have to get someone to upload them and get them running.  I would rather massage the data & inputs to get the best results possible, without resorting to code or version changes.  The rest of my layers 
are ready to run, and I want to start writing as soon as possible so that I might graduate in January/February rather than spread this agony out much longer.
	I have some experience with programming, but no C experience, and I am just learning to think in UNIX.  If you think the fix might take some new code, I should be able to get help in uploading and recompiling.

							Thanks for your assistance,
							Bryon Haney
							Novice user, soon looking for work

Please feel free to call me at 817-921-7270 or 7273, or return mail to me.  I am usually in from 9-7 or later central time, although I will be spending some time on field verification  of landuse  (and checking to see if my current results have any basis
 in reality) over the next week or so.  Thanks, bmh