L3DT users' community
Large 3D terrain generator

Mesas and sinkholes?

Got a problem or need advice?

Mesas and sinkholes?

Postby nicethugbert » Tue May 04, 2010 3:39 pm

How can I get more mesas and fewer sinkholes? The sinkholes are making my terrain impassable. It seems that the noise strength produces as many sinkholes as ... uh ... not sinkholes or mesas or tall stuff like that or stuff that is tall as sinkholes are sink.

EDIT 0: Alright, I'm gettign good results with 20%, 30%, and 40% erosion. My noise strength is 30%. I wonder if there is a relationship to be maintained between Noise Strength and Erosion if I want to keep the sinkhole count low as I increase the noise strength. 30% Erosion seemed the best. The cliff/terrace is maxed.
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm

Postby Telarus » Wed May 05, 2010 7:16 am

Use the setting that you've tuned to that sweet spot, then use Operations > Heightfield > Change vertical range... to adjust the terrain up or down relative to sea level/lake bodies. You can also do things like scale towards/away from 0 (scale the negative side only to flatten drops), or other interesting math stuffs.
Telarus
Doyen
 
Posts: 112
Joined: Mon Jun 01, 2009 12:34 am

Postby nicethugbert » Wed May 05, 2010 12:04 pm

Scaling the whole map will not get rid of the sinkholes.

Adjusting relative to a fixed point only works if your map is flat. Adjusting relative to water bodies only works if your map has them. Mine are not flat. They have slope and sizable features. The sinkholes are often dry.

The sink holes and their opposites are clearly a result of the noise algorithm. I suspect the noise algorithm is local height variance for each DM pixel and that it has a positive and negative swing. If we could control the polarity of the swing then we could control the presence of sinkholes and their opposites.
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm

Postby Aaron » Wed May 05, 2010 12:53 pm

Hi nicethugbert,

The sink holes and their opposites are clearly a result of the noise algorithm.


...and the cliffs / terraces algorithm, of course. If you're getting sinkholes in areas you don't want them, perhaps you could try reducing the cliffs / terraces values in those areas.

I suspect the noise algorithm is local height variance for each DM pixel and that it has a positive and negative swing.


Correct.

If we could control the polarity of the swing then we could control the presence of sinkholes and their opposites.


Slight problem. If the polarity of the noise was either all positive or all negative (depending on preference), then the resultant altitude of the heightmap at a pixel would drift away from the height set in the design map pixel. As it is, the noise is randomly positive or negative, which keeps the average height around about where you set it in the design map.

Have you tried removing the sinkholes using the 3D heightfield editor? I'm sure the leveller and erosion tools would fill them in.

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

Postby nicethugbert » Wed May 05, 2010 2:18 pm

I've tried using the editors but it's horrible because I'm too ham handed for editors.

It's never a simple thing to use an editor. The small details look out of place when gross adjustments are made. That's one reason why I wanted randomness features in the brush.

What L3DT does on it's own looks a trillion times more correct than what I can brush on, at least with simple brushes.

But, the algorithms have certain propensities.

It's true that if you clamp or shift the noise product then you change the average height, mathematically, but, not necessarily so to the human eye. If you clamp or diminish the negative swing, the human eye sees more average ground. Average height to the human eye is the height minus the extremes, not the extremes redistributed as a mathematical average would amount to.
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm

Postby Aaron » Wed May 05, 2010 10:43 pm

Hi Nicethugbert,

The magnitude of noise is such that I think the human eye really rather will notice the difference. Putting my...er..code...where my mouth is, I've run some tests:

Firstly, here is a 'normal' heightfield, generated using default settings, but with 100% cliffs/terraces. Here the peak noise is both negative and positive, so that the average is zero (i.e. noise does not change the average altitude, just makes it bumpy).

Image

Here is what happens when the noise is made all positive (same camera position):

Image

...and all negative (again, same camera pos):

Image

To be clear, the second and third images were made using the exact same design map and settings. The only difference was that I modified the noise peak generator to be either all positive or all negative, and removed the line of code that forced the sum of the noise to be zero.

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

Postby nicethugbert » Thu May 06, 2010 12:01 am

I can actually work with the all positive noise! That looks awesome! It's so Moria!

Of course it'll need to be adjustable.

What's the altitude range on your normal map? The settings seem highly sensitive to the size of the map in meters.

EDIT 0: I have the suspicion that it's not on the scale I can use for NWN2. On the scales I need for NWN2, sinkholes when present tend to dominate their surroundings. I notice that you have room to get around sinkholes in your first map. Also, your hills are rather low.

I have some maps I did for NWN2 2 years ago here. It was from my first attempts so hopefully I do better than that now. My current focus is certainly different.
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm

Postby Aaron » Thu May 06, 2010 1:46 pm

Hi Nicethugbert,



nicethugbert wrote:
I can actually work with the all positive noise! That looks awesome! It's so Moria!





The problem, however, was that the design map was decidedly un-Moria. I had a basic design map, which specified the maximum mountain height to be around 1000m (using 10m horiz. scale, BTW). Here's what it looked like:



Image



Using the regular algorithm (with positive and negative noise) gave me exactly what I (or other users) would have expected from that design map; a terrain with the same shape, and mountains up to 1000m:

[texture in 2D]
Image

[textured in 3D]
Image

Using the positive-only noise, I got instead a ridiculously mountainous terrain (mostly above the snow line), with mountains up to 5000m! That's not what I had in the design map, and I think most users would be dissatisfied to receive terrain that is so very different from their design.

[texture in 2D]
Image

[textured in 3D]
Image

Noise should not change the terrain to this degree. Hence, I do not intend to release this modification, or indeed any modification that has the noise producing a net change in elevation away from that set by the user in the design map.

The same 'Moria' effect could have (and should have) been achieved using the 'altitude range' control in the design map parameters wizard. That's what its for.

To avoid the mesas and sink-holes, the recommended method is still to reduce the cliffs & terraces setting, either for the whole map (in the design map parameters wizard), or locally (in the design map brush or pixel editor). And, if that fails, use the heightfield editor. A change to the noise generator will not be a robust solution to the problem of mesas/sinkholes.

nicethugbert wrote:The settings seem highly sensitive to the size of the map in meters.


The degree of sensitivity depends on the horizontal scale itself. The altitude range and noise amplitude both scale according to 1-exp(-ax), where x is the horizontal scale and a is an arbitrary scaling factor. Thus, both the altitude range and the noise amplitude increase with increasing horizontal scale, but asymptote towards a maximum value.

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

Postby nicethugbert » Thu May 06, 2010 10:20 pm

Wow, that Super Sized map looks awesome!

I agree that the results should not vary radically from the design map, unless the user wants that. But, the reason the average height is affected so radically by the local height is because of the equation or how it's being used. Another equation would produce an acceptable result, or, simply scaling the input to the current local height equation when it is used to produce an imbalance of positive or negative values.

L3DT has a tendency, under certain settings, to produce pentagon shaped depressions in quantity. This is great when you want it too. Not so great when you don't. Unfortunately, these depressions seem to be tied to positive swings which produce mesas and other local elevations above the local average. As you increase the cliff-terrace slider the depressions become less hexagonal, more rounded. But, the sides become impassable. This is all great if that's what you are going for.

But, If you want a more typical earth like terrain then you don't want a balance between local depressions and local protrusions. From what I have seen, and for what I need in game, local protrusions dominate over local depressions, particularly in a mesa environment. Of course, there are places where this is not so. But that is not the effect I am having trouble achieving. It's the mesa studded landscape with trespassable depressions that I am aiming for.

I am as we speak, running a series of experiments with L3DT to see if I can achieve that with L3DT as it is. I'm going to post them with the next day or 2 as I import them into NWN2 and make them available.
nicethugbert
Doyen
 
Posts: 132
Joined: Sat Feb 02, 2008 12:58 pm


Return to Help and support

Who is online

Users browsing this forum: Majestic-12 [Bot] and 9 guests

cron