Examples of using the Erosion Operator

Matt · February 01, 2010, 10:51:37 PM

The default settings produce quite old, heavily eroded terrain where valleys slump to produce larger erosion patterns. To create thin channels you can reduce the diffusion parameter. Increasing flow distance while reducing duration might also give results more like the GC thin flows if your terrain is quite large. However I concentrated more on long term erosion effects that simulate a mixture of short bouts of heavy erosion with long term subtle erosion (and anything in between), and that mixture can't be controlled directly.

"Difference (Erosion Field)" mode is intended for use as a subtle displacement which you might apply to a procedural landscape, so that you can keep the procedural details while calculating the erosion on a heightfield generated from that same procedural. To do that you would want to disable "flatten first" on the heightfield shader so that its displacement is additive.

CCC · February 02, 2010, 02:55:53 AM

Alright, Matt. I took your suggestions and i must add i am impressed with the results. Added to a Perlin Rigid fractal. There is a lot of variations of erosion going on both on a large-scale basis and small-scale. There are large cuts and tiny winding cuts, nice sedimentary deposits, even inverted flows and other shifts. This is a very nice erosion algorithm overall.

reck · February 02, 2010, 04:02:25 AM

CCC, that last image is very nice. For some reason this is a part of TG2 we don't hear a lot about. Glad to see someone taking the time to look into it.

EoinArmstrong · February 02, 2010, 04:12:38 AM

Oh now that's what I'm more interested in

Looks great - looks very WM2. What size hightfield was it, and how long did it take to calc (roughly)?

CCC · February 02, 2010, 04:21:57 AM

Quote from: reck on February 02, 2010, 04:02:25 AM
CCC, that last image is very nice. For some reason this is a part of TG2 we don't hear a lot about. Glad to see someone taking the time to look into it.

Indeed.

CCC · February 02, 2010, 04:25:42 AM

Quote from: EoinArmstrong on February 02, 2010, 04:12:38 AM
Oh now that's what I'm more interested in

Looks great - looks very WM2. What size hightfield was it, and how long did it take to calc (roughly)?

I just used the default 1000 by 1000 pixels/10,000 by 10,000 meters size for this one. Erosion on my Pentium 4 took less then ten minutes to calculate.

domdib · February 02, 2010, 04:57:22 AM

Very interesting experiments!

Tangled-Universe · February 02, 2010, 06:14:07 AM

Very interesting! Thanks for putting effort into exploring this territory.
I really like the last test you made. Like Eoin said it looks very similar to WM2, which is good.
The calculation time isn't bad either!

Cheers,
Martin

CCC · February 02, 2010, 07:57:34 PM

I will now be creating a preset collection of Erosions at NWDA. It will be some time as to the complex depth i will be going into investigating these features, however once the collections are done, there will be an opportunity for users to add eroded types of terrains to there local scenes.

Stand by!

choronr · February 02, 2010, 08:18:10 PM

Quote from: CCC on February 02, 2010, 07:57:34 PM
I will now be creating a preset collection of Erosions at NWDA. It will be some time as to the complex depth i will be going into investigating these features, however once the collections are done, there will be an opportunity for users to add eroded types of terrains to there local scenes.

Stand by!

Thank you; and, we'll be looking forward.

Gannaingh · February 03, 2010, 01:53:50 AM

That last one looks awesome! Add some textures, some trees, and some clouds and you have the makings of quite a realistic scene. THank you for the hard work you've put into this, the results are awesome!

Tangled-Universe · February 03, 2010, 02:08:22 AM

Quote from: darthvader1 on February 03, 2010, 01:53:50 AM
That last one looks awesome! Add some textures, some trees, and some clouds and you have the makings of quite a realistic scene. THank you for the hard work you've put into this, the results are awesome!

Speaking about masking:

In theory you should be able to generate color from the erosion. I'm just thinking out loud here, so I might be mistaken on how to do it exactly, but I believe you can. If you load the terrain twice in seperate heightfields you could use the eroded one for your scene.
I'm not sure if the heightfield-shader can generate color. If so, then you could subtract the eroded heightfield from the original and use the subtracted color as a mask for the eroded parts of the terrain.
Very nice idea btw of making a pack of different erosion-patterns. If you could get my idea described above get to work it would be really awesome

Cheers,
Martin

Henry Blewer · February 03, 2010, 08:52:22 AM

These new tests look very promising. I really think the last render looks great. Generally I used simple power fractals to simulate erosion with the y axis stretched. Various scales provide the variation of older an newer erosion. But I don't go too far, my Pentium 4 bogs down when heavy calculations are needed.

domdib · February 09, 2010, 07:36:44 AM

With my limited testing of the erosion operator, it seems to be only single-threaded. It would be awesome if it could be multi-threaded.

Also, the advice Matt gave is very useful. Does advice like this also apply to the Alpine fractal, which seems to be simulating erosion? If so, it might be nice to have a few typical examples of specific sets of settings.

@CCC - just to be clear, in the last example, are you applying the "Difference (Erosion Field)" to a Heightfield Generate with Perlin Ridges?