Author Topic: Research papers on erosion/rivers... Possible feature request?  (Read 5711 times)

0 Members and 1 Guest are viewing this topic.

Tangled-Universe

  • World Machine Guru
  • ****
  • Posts: 66
    • View Profile
Hi guys,

I ran into these 2 interesting research works on erosion simulation:

http://hpcg.purdue.edu/pubs/2005/Benes05EGNAT.pdf

In WM I find it difficult to get large talus depositions as like with these table mountains, but also like in terrains like these:
http://www.nextgen-gallery.com/wp-content/gallery/landscapes/landscape_09.jpg
I really like the way their method simulates talus deposition.
It seems WM isn't really doing the "parallel" type of erosion where the eroded cliff face keeps its steepness and just regresses on the horizontal axis, thus mainting its steepness and much of its height.
In WM the erosion tends to do it the other way around, by eroding away the surfaces, effectively flattening it out mostly.
This might be an incorrect observation and user error of mine, so I'm happy to hear about this anyway :)

The paper below is rather a complete terrain development solution which we don't need, but it utilizes an interesting river system, which I seem to remember a feature request of here as well:

http://arches.liris.cnrs.fr/publications/articles/SIGGRAPH2013_PCG_Terrain.pdf

Perhaps you are aware of these papers already Stephen, but I thought I'd just bring them to your attention.

I'm curious to hear what you think about it, especially about the erosion.

Cheers,
Martin


Remnant

  • Author
  • World Machine Guru
  • *****
  • Posts: 2311
    • View Profile
    • http://www.world-machine.com/
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #1 on: May 14, 2014, 12:54:47 PM »
Hi Martin,

Thanks for bringing these papers to my attention. I have not read either before -- my thoughts below after a quick read:

The first paper is describing an algorithm that is essentially identical in practical aspects with World Machine's Thermal Weathering in "Two Phase" mode. This method models the accumulation of talus and scree as a separate type of material from the underlying rock, that has a different (lower) angle of repose. This allows debris to form and flow away from cliff faces while keeping the cliffs more or less intact. You should be able to get quite nice table mountain using this mode!

The terrain from rivers paper is quite interesting -- I will take a more in depth look later. There are some exciting things coming up soon for World Machine with regards to rivers though, I may mention :)
-- Stephen Schmitt
- Creator of World Machine

Tangled-Universe

  • World Machine Guru
  • ****
  • Posts: 66
    • View Profile
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #2 on: May 15, 2014, 05:47:22 AM »
Thank you for your swift response Stephen :)
I'm glad to hear you find it interesting, if you don't mind I may post more in the future in this topic.

I'm excited to hear about the upcoming river-related stuff in WM, looking forward to it!
Is there any (rough) date for release announced?

It seems I just need to practice a bit more then with the Thermal Erosion device and especially with the angle settings.
If I can't get it to work to my liking then I may bug you again :)

Cheers,
Martin

Remnant

  • Author
  • World Machine Guru
  • *****
  • Posts: 2311
    • View Profile
    • http://www.world-machine.com/
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #3 on: May 15, 2014, 11:32:19 AM »

The biggest problem with the thermal erosion device is that its results can tend to be either too pronounced or maddeningly subtle -- so for general earth shaping it serves in more of a secondary role. There are a variety of things I want to do to improve that. One thing is just improving the tweaking parameters (for example, extending the duration range so that you can run longer simulations), while it could also valuably use a multigrid simulation like the "Geological Time Enhancement" option of the erosion sim.

One other thing I noticed is that in that paper they intentionally randomize the amount of thermal erosion at each location on each timestep to produce a rough surface. This could be a valuable addition to WM's thermal erosion, as it produces very smooth surfaces which are not always what you want.
-- Stephen Schmitt
- Creator of World Machine

Troedel

  • World Machine Veteran
  • ***
  • Posts: 25
    • View Profile
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #4 on: July 24, 2014, 01:26:00 PM »
Quote
The terrain from rivers paper is quite interesting -- I will take a more in depth look later. There are some exciting things coming up soon for World Machine with regards to rivers though, I may mention :)

Any news on this? Creation of proper drainage is very hard so any help would be welcome ;)

Greetings,

Troedel

Remnant

  • Author
  • World Machine Guru
  • *****
  • Posts: 2311
    • View Profile
    • http://www.world-machine.com/
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #5 on: July 24, 2014, 05:19:27 PM »
Yes!

No imminent news, but there is a major feature being implemented that will be surfacing sometime this fall involving synthetic river modeling. I will be posting something on the blog about this sooner or later, as it is an especially exciting development for a number of reasons. Just the teaser about it for now though!


-- Stephen Schmitt
- Creator of World Machine

Troedel

  • World Machine Veteran
  • ***
  • Posts: 25
    • View Profile
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #6 on: July 24, 2014, 11:34:01 PM »
Thank you for the rapid response. Looking foreward to it ;)

Tangled-Universe

  • World Machine Guru
  • ****
  • Posts: 66
    • View Profile
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #7 on: August 18, 2014, 08:11:02 AM »
Hi Stephen,

I was wondering whether the "randomize of the amount of thermal erosion at each location on each timestep to produce a rough surface" + other possible improvements to the Thermal Erosion devices will make it into WM3? Or just one of the mentioned aspect?
Just curious :)

The river stuff sounds very exciting too!
Is this something you are developing yourself completely or do you use existing research (papers) similar to the ones I posted here?
I'm curious to see and read about the model, if it's out there of course.

Cheers,
Martin

Remnant

  • Author
  • World Machine Guru
  • *****
  • Posts: 2311
    • View Profile
    • http://www.world-machine.com/
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #8 on: August 18, 2014, 02:32:17 PM »
Thermal erosion improvements will not making it into dev-1, but they are competing for attention for later 3.0 dev releases.

It's too soon to start talking about the river modelling, but yes; it involves collaboration with some good folks in academia to implement their river model in WM.
-- Stephen Schmitt
- Creator of World Machine

Tangled-Universe

  • World Machine Guru
  • ****
  • Posts: 66
    • View Profile
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #9 on: August 26, 2014, 03:39:13 AM »
Thanks for getting back to me about this Stephen.

Regarding the river modelling, this sounds interesting as well as a wise approach.
Better to not try to invent the wheel yourself, if there are academic resources (paper or people) available.

Regarding the erosion models:

I'm not sure how much it is competing for attention for later dev releases, but may be an interesting incentive to improve WM's erosion further is because I understood yesterday from a guy who used to work at Framestore that their choice for terrain modeling is GeoControl instead of WM and the reason for that was primarily because they liked GC's erosion much better. Secondly, but of minor importance compared to the erosion, it is also available for Mac.
They used GC in Maleficent (opening scenes show it) and now they have also used it in Guardians of the Galaxy as a consequence.

WM is definitely on the radar as it seems, but loses to an imho inferior product mostly because of erosion simulation. That's a little bit of a shame imho?
So that's some light shed from the vfx industry.

From my perspective WM has matured greatly and has improved very much on user-friendly tools for art-directive terrain modeling in the past years.
I can now produce basically any base terrain and/or base terrain layout shape for a landscape I like. The tools are there, great!
But if I ask myself "Can I make more realistic and nicer looking terrains with all this control?". Not really much more to be honest, not that much more compared to how much more control I have now, if you see what I mean.
If I wish to make more alpine-like terrains then the erosion settings I need too often result in either recognisable squarish patterns and/or with deep sharp straight channels (inverse filter, mostly).
I hope development will catch up a bit on that in the coming years to even out the attention areas a little.

I'm not trying to put you on a spot here with the industry example and my thoughts/hopes for the future, but I hope you may appreciate and understand my increased interest in this area a bit better now, also from my user perspective :)

Thanks!

Martin

WFab

  • World Machine Guru
  • ****
  • Posts: 402
    • View Profile
    • Blog
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #10 on: August 26, 2014, 04:02:01 AM »
Thanks for bringing up that point about straight erosion patterns. If that problem is fixed then there is no match for world machine in terrain modelling market. Although we can distort it a bit with simple displacement filter, it still looks somewhat synthetic. Someone on this forum once suggested an alternate model (using cellular automata I believe) for realistic erosion. I don't know from a programmer's perspective how hard this change would be, but I would really like to see this in my lifetime  :P

About the new "in the pipeline" river model, would it be able to simulate some realistic glacial erosion?

Remnant

  • Author
  • World Machine Guru
  • *****
  • Posts: 2311
    • View Profile
    • http://www.world-machine.com/
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #11 on: August 26, 2014, 10:39:49 AM »
Hi,

I will try to respond more fully later, but just thought I would mention one thing while I'm here:

If you've tried the new 3.0.dev1, try changing the reconstruction type from Linear to Smooth. This changes some of the character of the erosion, and usually removes axis-aligned ridge/valley artifacts from the erosion results.
-- Stephen Schmitt
- Creator of World Machine

Tangled-Universe

  • World Machine Guru
  • ****
  • Posts: 66
    • View Profile
Re: Research papers on erosion/rivers... Possible feature request?
« Reply #12 on: November 02, 2014, 07:06:15 AM »
Hi Stephen,

Hi,

I will try to respond more fully later, but just thought I would mention one thing while I'm here:

If you've tried the new 3.0.dev1, try changing the reconstruction type from Linear to Smooth. This changes some of the character of the erosion, and usually removes axis-aligned ridge/valley artifacts from the erosion results.


I'm still curious to hear what else you had to say in regard to this topic :)
The smooth reconstruction type sometimes produces more pleasing results, but usually I need to increase the resolution an extra notch, because the linear interpolation produces more detail often.

Would you have time to take a look at the thermal erosion deposition randomization we discussed before? I might be heavily mistaking, but it doesn't sound overly complex to try and as you may understand by now I'm extremely curious :D

FYI, I just found this paper which you may find interesting to read:
http://www.cs.rpi.edu/~cutler/classes/advancedgraphics/S09/final_projects/odhner_fleming.pdf
No feature request following this paper, just to share with you ;)

Thanks for your time!

Cheers,
Martin