Aspect map generation

[Charles Ehlschlaeger]

In article <memo.654238 at cix.compulink.co.uk>, nra at cix.compulink.co.uk (N R A) writes:
|> 
|> We've been using the 'r.water.outlet' program to generate catchments. The 
|> aspect/drainage map for this program was generated by 'r.watershed'. 
|> Rather than keep regenerating the aspect map each time, I decided to do 
|> it for our whole region. I created a shell script to create a 10 km 
|> squares and then 'stitched' these together to create the single big map. 
|> I was concerned over any 'edge' effects, so made the 10 km squares 
|> slightly bigger (10 km + 100m for a 50 m resolution DTM) and then cut out 
|> the centre 10 km.
|> 
|> Everything seemed OK, derived catchments look fine, but now find that 
|> there are some weird effects which seem to be on the boundaries of the 10 
|> km squares. Basically the derived catchment didn't include some rivers, 
|> which it should.
|> 
|> If I zoom in to cover the catchment, generate another aspect and then the 
|> catchment, the missing rivers get included. I've done a 'r.infer' and 
|> found the differences - along the 10 km grid line.
|> 
|> Question is - does anyone know how the aspect part of 'r.watershed' 
|> works. Primarily is there an 'edge' effect and how far does it extend - 
|> on the test above I could see difference around most of the edge. Some 
|> extended upto maybe 10 cells into the map.

Yes, there is an edge effect, but it shouldn't extend more than 2 or 3 cells
beyond the edge of the map.  Try again with a 300m overlap instead of 100m.
If you want to experiment to determine the minumim overlay, r.infer several
edges against regions with no edges nearby.  That should show you if some
sections of the map have greater edge effect than others.  Choose an overlay
several cells larger than the widest overlay.  Better safe than sorry. 
Let me know your results.

|> Or better yet how can I generate a single aspect map from our DTM map in 
|> one go ? - now that would be really useful  8-)

Buy more RAM, add more swap space, and leave r.watershed running for days on
end 8-)

chuck