Posted by Jerry Stuckle on 02/16/07 15:33

Toby A Inkster wrote:
> Jerry Stuckle wrote:
>
>> Nice algorithm. For your dead end, might I suggest you make the lowest
>> square a lake and take the next lowest square as the outlet? Repeat
>> until you get an outlet.
>
> If I understand you correctly, I think you might end up with lakes that go
> uphill that way. Consider the following landscape; heights in
> metres, co-ordinates as A1, C2, etc:
>
> A B C D E F G
> 1 800 600 400 500 600 800 800
> 2 800 550 350 480 550 650 600
> 3 800 550 200 300 310 300 300
> 4 880 650 220 400 400 450 500
> 5 900 700 240 400 450 500 550
> 6 900 700 280 450 500 550 600
> 7 900 650 350 500 550 600 650
> 8 900 650 300 550 600 650 700
>
> Now imagine our river comes onto this map from the North at C1. Using my
> algorithm it flows to C2 and then C3. Here it reaches a local minimum, so
> should form a lake.
>
> Using your algorithm, it will look to C4 as an outlet. Assuming we can't
> go back to C3 as this would result in an infinite loop, we are again at a
> local minimum, so we look to C5 as an outlet, and so on, eventually
> expanding the lake into C5, C6 and C7 and then finally finding a way
> downhill at C8.
>
> In real life, although C3, C4, C5 and C6 would fill up as a lake, C7 would
> not; D3 would become part of the lake, and the outlet would be through E3,
> F3 and G3.
>
> It's a rather contrived map -- a valley with two possible outlets for the
> river: one that initially seems less favourable, but ultimately should
> prove the chosen outlet. I'm sure it would crop up from time to time
> though.
>
> It *can* be dealt with computationally, but it's rather hard.
>
> Personally I'd be tempted to not give the lake an outlet and explain it
> away by saying that the water flows into an underground stream from there.
> This might be a realistic explanation for one or two lakes on a map, but
> not for hundreds though -- having not run the algorithm myself, I'm not
> sure how often they would crop up.
>

Tony,

Yep, that's just what I'm saying. And if the lake gets too big, you
could adjust the elevation a little to bring it down.

The underground stream is also a possibility. Not sure how it would go
over, though. It does seem a bit contrived. Besides, rivers are a good
source of transportation, food, etc.

Another way I thought about doing this is to make it more realistic.
Instead of having just a single elevation for each square (which isn't
realistic in most places), have a minimum and maximum elevation. That
range could increase as the minimum elevation increases (i.e
foothills/mountains) and decrease as the elevation decreases (plains,
etc). It might give a more realistic pattern, and more possible outlets
for the lake.

And small changes to the algorithms could have large effects on the outcome.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
jstucklex@attglobal.net
==================

[Back to original message]