L3DT users' community
Large 3D terrain generator

Sinkholes

It doesn't hurt to ask...

Sinkholes

Postby nicethugbert » Tue May 11, 2010 6:10 pm

Sinkholes are primarily a result of underground water movement. Very rarely are they a result of above ground water action. In such cases it is typically because of human activity, man made ponds that weaken the ground they rest on because of their weight and their water action combined.

Cliffs are usually a result of weak rock being eroded or weathered away from rock that is highly resistant to such or by ground faults.

Terraces are a result of above ground water action.

L3DT produces sinkholes as a result of the interaction of Cliff/Terracing and Noise Strength, not underground water activity. As a result, sinkholes are always strongly coupled to cliffs/terraces in L3DT unlike in the real world where they can be highly independent.

The result is that there is poor sinkhole control in L3DT. Can something be done to address this?
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm

Postby Aaron » Tue May 11, 2010 10:19 pm

Sure, but I'd have to remove the existing cliffs / terraces algorithm. The replacement, to be accurate, would need to be a full geophysics simulation to model long-term erosion and cracking of rock strata, taking into account different rock hardnesses, permeability, soil transport, etc. That's not a small amount of work. Optimistically, it might take years to get right, assuming I have the required skills to build such a simulator. If I did this, I could guarantee L3DT licenses would not be selling for $35 any more; zero(s) would be appended.

Edit: for that matter, you could always download the plugin API and build your own more realistic cliffs / terrace effect. I'm sure it would be difficult, but no-doubt rewarding.

...or we can accept that the quick-and-nasty (and cheap!) cliffs/terraces routine sometimes produces more sinkholes than one might think it should, and perhaps practice using the design map brush to prevent sinkhole formation or even the heightfield tools to remove them afterwards, as I have suggested in several other threads on more-or-less this topic.
User avatar
Aaron
Site Admin
 
Posts: 3696
Joined: Sun Nov 20, 2005 2:41 pm
Location: Melbourne, Australia

Postby nicethugbert » Wed May 12, 2010 5:36 pm

Could you append the zeros to the front instead of the back? Hehheh, just bustin.

There has to be a simpler way around this problem than full Geo stuff or Brushes. I'll be old, gray, and spewing death bed confessions before I finally develop brush skills. I'm just really really behind the curve on anything artistic, or expensive, and so are most people. So, I think there is a huge market for the tools I need.

And, sorry, I did not realize that I was back to the same issue. Now I think that the sinkholes/mesas problem is related to the cavern and tunnel system problem.

In not to overwhelming technical terms, what is the technical nature of this problem? Is this an issue of local variance? Can we have some local variance control type stuff? Noise Strength is Noise Amplitude? Is noise strength both the frequency and the amplitude of the noise equation?

EDIT 0: How local is local? What's the radius or unit measure of the noise equation, or other equations?
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm

Postby Aaron » Thu May 13, 2010 1:06 pm

Hi Nicethugbert,

...and so are most people. So, I think there is a huge market for the tools I need.


I'm yet to be convinced, clearly. If there were a huge market, I would have expected someone else to have requested these tools by now. L3DT has been publicly available for seven years. In that time, some of the most requested and popular features, aside from faster calculations and yet more file formats, have been the more and better brush tools and editors. Most users may not be artists*, but certainly a great many users want and use the editor tools. Maybe not every day, maybe not because they like using brushes, maybe not even because the L3DT tools are nice or easy to use, but because they get the job done.

* that's also debatable; based on my correspondence with users, I'd say most are a blend of developer/artist/modder. That sort of multi-skilling is necessary for indie developers, which make up the majority of L3DT licensees.

nicethugbert wrote:And, sorry, I did not realize that I was back to the same issue. Now I think that the sinkholes/mesas problem is related to the cavern and tunnel system problem.


Er...I was referring to your other mesas and sinkholes thread.

nicethugbert wrote:In not to overwhelming technical terms, what is the technical nature of this problem?


As I think I explained in the other thread, the problem is that the map contains some depressions (either from noise, from the design map altitude layer, or erosion, or whatever), and the terraces algorithm is converting the depression edges into cliffs. So, you need to prevent depressions forming, filter them out, or change where the cliffs/terraces are applied.

I'm not convinced of these objectives could be accomplished by changing the noise algorithm or giving you more control over it. You can get sinkholes on maps with no noise. Of course, you do get more on maps with noise, because noise produces peaks and depressions. However, even if the noise only added positive peaks, some spaces between the peaks would still be depressions, and they'd result in sinkholes when terraced.

I don't believe the problem of sinkholes will be fixed by any change to the noise algorithm. Taking ten minutes to use the raise, smooth and perlin brushes in the 3D editor will (in considerably less time, I might add, than it has taken us to debate the issue). Any such change to the noise algorithm will, however, make the design map more difficult and confusing to edit, slower to generate the heightfield, it'll require more documentation, and be more difficult to maintain, not to mention consume development time in messing around to find a suitable noise algorithm (should one exist). The cost vs. benefit ratio is, to me, poor. There are better prospects for the use of my (sadly scarce) development time.

Anyway, we've been through this in the other thread. I'm not convinced, but you're perfectly welcome to try. You may be able to get the effect you want by making a different noise algorithm in ZeoScript (slower) or by using a plugin (faster, needs MSVC), but you'll have to judge that cost vs. benefit ratio for yourself.

nicethugbert wrote:Is this an issue of local variance?


Yes, in part.

nicethugbert wrote:Can we have some local variance control type stuff?


'Feature scale' and 'Noise strength' principally determine local variation in height. For more localised local control, there are [cue broken record] brush tools.

nicethugbert wrote:Noise Strength is Noise Amplitude?


Yes.

nicethugbert wrote:Is noise strength both the frequency and the amplitude of the noise equation?


No. The range and proportion of spatial noise frequencies is invariant with the noise amplitude. Higher frequency noise just makes the map look jagged, and lower frequency noise usurps the role of the elevation layer in the design map. Noise fills in that middle ground.

nicethugbert wrote:EDIT 0: How local is local? What's the radius or unit measure of the noise equation, or other equations?


For peak noise, using the 64x ratio for HF/DM, the distribution of peak radii works out as being is between 12.8 and 153 output heightfield pixels. Of greater significance than the limits are that the distribution is skewed heavily towards the lower-frequency peaks (#peaks is inversely proportional to the square of the peak radius).

For fractal noise, it's not possible to define a radius or distance metric. Fractal noise is present at all scales.

...and, saving the first for last:

nicethugbert wrote:I'll be old, gray, and spewing death bed confessions before I finally develop brush skills


That just comes across as defeatist. I'm no artist either, and I don't like using manual brush tools, but when I need to use an editor, I do, or else the work just doesn't get done. Really, what's the alternative?

Regards,
Aaron.
User avatar
Aaron
Site Admin
 
Posts: 3696
Joined: Sun Nov 20, 2005 2:41 pm
Location: Melbourne, Australia


Return to Feature requests

Who is online

Users browsing this forum: No registered users and 6 guests

cron