[GRASS-dev] r.watershed updates

Helena Mitasova hmitaso at unity.ncsu.edu
Sun Oct 29 18:06:46 EST 2006

On Oct 29, 2006, at 4:22 AM, Hamish wrote:

> Hi,
> Now the lots-of-colors bug is fixed, I've now run r.watershed on a  
> large
> grid (5000x5000 cells [20% null]). It took 54 hours, basin/half-basin
> results look nice, especially with: "nviz elev=topo color=basins.half"

Hi Chuck - it is great to see you involved again. Thanks for looking at
the color problem - it was a long standing issue that you were the  
best person
to resolve. We use r.watershed a lot with lidar and SRTM data and it  
very well.
> r.terraflow with the same area & mem=1500MB took just minutes, but  
> only
> outputs sink-watershed basins, not all basins. (I may be missing
> something here, this is all outside my field)

r.terraflow is not designed to output basins the way r.watershed does.
there is another code that I hope will be eventually contributed that  
terraflow output and creates a map of watershed hierarchy.
> some weirdness/comments:
> Accumulation map: number of cells that drain through each cell
>  |   Range of data:    min = -16269552  max = 8562412
> (way negative? way postive?) bulk of data is 0->2000 though, so after
> setting color rules the map looks ok.

Did you read the manual?
> stream map: 0 -> null
> basins,half-basins: 0 -> null
> length-slope, slope-steepness maps show a lot of cross-hatching
> artifacts. (dem is r.contour output; slope, pcurv, tcurv look alright)

I am not sure what you mean here, but I think that is because of the D8
algorithm. For LS factor that should be computed only for hillslopes
(becuase that is what USLE/RUSLE is for)
r.flow works better with Dinf method.
  But for stream networks it is hard to beat r.watershed.
(Chuck you should write a full paper about it - I get
lot of questions about the algorithm and where it is published that I  
cannot answer,
the proceedings paper does not give too much detail.
If you would like some good applications, I have plenty.)

> with such a long processing time the iterative process of picking the
> correct threshold size requires a bit of planning.
> used ~ 570mb RAM (all-in-memory mode)
> Hamish
> _______________________________________________
> grass-dev mailing list
> grass-dev at grass.itc.it
> http://grass.itc.it/mailman/listinfo/grass-dev

More information about the grass-dev mailing list