Gaps in flow maps on flat terrain

Hi All

I use WM to generate masks for Terragen, the most useful being the flow map. For the most part the erosion algorithm runs beautifullly but it seems to run into problems when it reaches a level section of terrain.

Example image (115kb):
<http://www.path.unimelb.edu.au/~bernardk/demo/flow_example.jpg>
Full render (1.2Mb):
<http://www.path.unimelb.edu.au/~bernardk/demo/glacier.jpg>

The river banks (coloured dark green) in the top half of the image have a relatively consistent drop in altitude along their length while the river system in the bottom half of the image is in much flatter country resulting in lengthy gaps in the mask where the slope approaches 0°.

The other main problem occures when a river runs down a slope onto a flat sutface (e.g. lake surfaces from DEM data). In this case the erosion occurs around the edge of the flat surface rather than extending into it (as if the water has no momentum, but is driven by slope alone)

Does anyone know of a combination of filters I could use to pick out troughs in a terrain so I can fill these gaps in the masks? Or is there something in v1 that will help?

Hi Ben,

if I understand what the problem is, its that the river system effectively disappears when you hit the flats, correct?

As is, there’s unfortunately nothing to do about it – this is an unavoidable artifact of simulating erosion. Let me describe roughly how the WM erosion algorithm runs:

Effectively, erosion runs many small streams down the sides of the terrain. The water erodes and shapes the terrain; the river system output is simply a look at where the streams are most often at across all the thousands of runs.

With this in mind, it’s easy to tell what’s going on: The steep areas of terrain have very distinct low areas and places a river would run; but when you hit the flats, there no longer is any real place for the water to prefer to run. There’s no large mass of water, like in real life, that will bore out a single meandering channel; instead it will just pretty much pick its own path. This results in the flowmap more or less dying completely when it hits the flats.

How to fix this?

That’s a tough problem. If you could do something to the terrain to keep a slope running through the lowest places, the water would have a place to track. I think any solution is likely going to have to do something like this.

I’ve got to run out right now… anyone here have some ideas?

Thanks for the explanation of the algorithm. It helps in so far as I know what tests are not going to work.

It looks like I’ll be doing some manual editing… although I’ve figured out some masks which might help.

1: Flow map works best (for me) with minimum river strength and shape AND minimum world size… This minimises the gaps.

2: Multiply the landscape by itself, set world size to maximum and select minimum slope.

From there it’s off to Photoshop for some creative blending and manually deleting those flat regions that are not in “valleys”. It’s easier to delete than draw. It means 2 builds in WM but it’s loking like a promising compromise. I knew I’d find a use for all this RAM :wink:

Try this:

Since even all but the flattest terrain has some variations, pick out troublesome sections on a piece by piece basis and exaggerate the profile by 2-3 or more times so that it is steeper. Do your analysis, use the results as a mask, and overlay the results on the original (flattish) terrain. It may not be perfect, but it will indicate the trends in a better way than just the flattish surface would. Sort of a 2nd approximation.

Tim O

Hi Ben,

How to fix this?

That’s a tough problem. If you could do something to the terrain to keep a slope running through the lowest places, the water would have a place to track. I think any solution is likely going to have to do something like this.

I’ve got to run out right now… anyone here have some ideas?

Here’s some screen grabs of my masks in Photoshop prepared using the method in my previous post.

Flow maps only
<http://www.path.unimelb.edu.au/~bernardk/demo/flow.jpg>

Flow maps + flat terrain
<http://www.path.unimelb.edu.au/~bernardk/demo/flow+flat.jpg>

It’s worked pretty well on the moutainous region on the left. picking up lakes, flats and hanging valleys without picking out too many ridges. In the flatter terrain on the right a number of ridges have also been selected but these can be easily removed manually.

There are still a few situations where problems arise but I’ll try some renders first to see if they’re worth worrying about because it’s very hard to manually edit masks areas where the terrain doesn’t give you many clues as to where to go.