River Networks

[Criss]

Sense GeoControl uses a L-System method what would be nice is the ability to generate river systems that flow off of mountains and maybie glaciers that drain naturally into the lowest points be it a lake or larger body like a sea and perhaps a method of delta formations but i am aware that these are complex algorythms to write of which had been done only semi-sucessfully with MojoWorlds River Plugin but still neat idea if it can be implemented some day...... :)

[imported_cajomi]

In version 2 there will be a special river tool. I do not know Mojoworld, and so, I don't know, how it works there.
Also, another method of placing erosion with the brush will be implemented, to set the beginning of a flow, but without any limits, where the flow ends.

[hillrunner]

Hi Johannes,
Indeed Mojoworld has some cool rivers features. River plugin was written by James Bardeen (famous physician).
I've found these interesting links here :
http://www.graphicsinterface.org/pre1996/93-PrusinkiewiczHammel.pdf
http://online.itp.ucsb.edu/online/colloq/bardeen1/
Olivier

[Criss]

Oh very nice indeed... i like that River Tool idea! :D

Yes... James Bardeen... admire his work allthough i can't understand most of it, heh,heh... math was never my forte............

[imported_cajomi]

I have taken a fast look at model of Prusinkiewicz and Hammel, based on midpoint displacement (triangles not squares) and Mandelbrots squig algorithm.

Well, that would be a a terrain shape filter, with strong level use, all levels have to be checked.

But as the authors wrote, there are some problems with this model. Most the constant altitude and the asymetric valley form.
The constant altitude may be not a so great problem, it would refer to a low slope, but the asymetric valley are really bad. It is very good visible in the renders in the document, that the strong slope at one side of the river, nearly 90 degree, is very unnatural.

I am thinking about a more realistic modell, with more then a nice riverform. It is based not so much on fractal random, but on physical processes.
So, you see, that also will be a really innovative tool.

[monks]

hmm- this is a problem we're currrently trying to 'solve' at ME-DEM. As terrains get larger, rivers become more important because terrain models become geography as opposed to arbitrary terrains with rivers running 'off screen'.
I like the sound of the approach your taking with this Johannes: physical process as opposed to fractal.
Aren't the MojoWorld Rivers procedurally created at render time?- I didn't think they existed as part of the terrain model.

monks

[Criss]

Some rather intreasting information about terrain fractals......

https://www.internal.pandromeda.com/engineering/musgrave/unsecure/S01_Course_Notes.html

In the site link above read the part about Dilation Symmetry as it talks about how it relates to river networks... crazy stuff...... :shock:

---------------------------------------------------------------------------------------------------------------

The 2.04 version of Bardeen's Rivers plug-in enhances the capabilities of Version 2.0, particularly when creating large continents with many levels of adaptive river detail. Features include:

* enhanced erosion effects, using rivers to sculpt the landscape.
* improved channels at lake outlets.
* greater consistency in the river network and channel widths between different levels of river detail.
* automatic adjustment to memory limitations of the host computer.
* new sample river worlds, including some with continents thousands of kilometers across to explore, illustrating a wide range of the infinite variety of possible parameter settings.

The continent-scale river terrains are resolved with MojoWorld's pixel-level detail, complete with alpine lakes, small cascading brooks, meandering streams, and rivers on the order of the Mississippi or the Amazon, which spread out into wide estuaries as they approach the sea.

------------------------------------------------------------------------------------------------------------

In MojoWorld before you do the final render you select the River Plugin and use a perimeter bomb to drop the river network over the terrain or continent and it grabs or adapts to the elevation, ridges, slopes etc. It's crazy stuff and these rivers can span up to thousands of kilometers. Although this increases rendering time greatly so doing river in a pre-existing heightfield i think is more preferred anyways.

But whatever is brought to GeoControl 2's table we all look foward to it. The way i see it River Systems is a heaven sent feature for terrain heightfield modeling period.

[monks]

The 2.04 version of Bardeen's Rivers plug-in enhances the capabilities of Version 2.0,

-so could you export the Mojo terrain with the rivers in place to another app and have a heightfield with rivers?

monks

[Criss]

Yes. Sense any given mojoworld terrain can be exported. Rivers are automatically a part of the terrain at that point once the perimeter bomb sets them in. I see this one terragen artist Hillrunner do it all the time as he uses imported mojoworld terrains within terragen. I was going to try it myself but i only had the demo of mojoworld. :)

[Criss]

There might be a way to mask out the rivers as well but i am not to sure. I thought i had seen it done somewhere before? Still Johannes idea sounds very nice to say the least. :)

[monks]

Wow. Thanks Criss, I'll have to look into Mojo.

The problem is twofold as I see it. You want both the ability to create rivers by hand BEFORE erosion is applied and in a manner that the Mojoworld and Johannes' tool (to generate the extra detail of tributaries and also rivers that you simply haven't the time to create yourself).
As far as before erosion, your usual method is to use either a piantbrush or a mask yes? The problem is you cannot guarantee with a brush that the terrain along the river is constantly falling. GeoControl is extremely good at creating realistic ridge networks with the brush, but the problem is the valleys are not guaranteed to have the correct gradient. Even a mask might not correct certain incorrect gradients because a mask wil have a narrow range of displacement setting- it's very good for varying the width and depth, but the real difficulty is *targetting* the displacement at those trouble spots and getting the constant falling gradient.
I suppose you could argue that it's kind of academic to get a perfectly correct gradient since at the end of the day, the model ends up in a game engine or a renderer, and they generally take care of rivers in their own way anyway. But, obviously there is some call for it, as Mojoworld has shown.

My 'perfect' river setup would be the following::)
*A vector or spline tool to draw primary watercourses. You would have the ability to drape vector to terrain. You would be able to vary the stream width and depth all along its path via profiles or control points. World Machine is well on the way to having this.
*The same vector tool which actually determines the height of the terrain along it. you would draw it and then click on it to open a dilaog or whatever. You then specify its start and end height. Then you click a button which says 'Interpolate along its length' to give you the gradient. Better still would be a 2D graph on which yuo draw a profile of the gradient. When the terrain is generated you would have (a wierd looking) skeleton of terrain under the vectors set at the heights. You would then use a brush or whatever (terrain lofted splines) to fill in the terrain between the rivers. You would have a perfect terain at the end.That's the idea anyway!
You could then go onto to run erosion in the usual way. The channels created by the erosion would naturally link up with your major rivers to form natural tributaries. I also think that Johannes' tool of defining a point from which erosion runs is very cool and would compliment WM's vectors perfectly.

So with this setup you would do:
*Mark out your PRIMARY rivers manually in WM vectors.
* Use GC's point and click erosion tool to generate SECONDARY rivers. Would this possible?
*Run a standard erosion over all to get the complexity of TERTIARY rivers.

So you would get the best of both worlds. Manual placement and procedural speed for levels of detail.



Leading on from Criss's point about L system and rivers,.another possible way of generating terrain is from the rivers. Generate your rivers first and then the terrain in between. I know that this has been covered before a little (SIGGRAPH apparently) but never really been developed.If the river generation is correct (gradients are correct), then this would ensure that the terrain is correct.

If GeoControl had a DLG river mode for its Brush, River Brush. that would be awesome. You would be painting an L system of watercourses. You would paint your River L system setting the terain protection slider as usual. This would influence the river network, complexity. The filters needn't be any different from the Terrain Filters, because the rivers and ridges are really just two sides of the same coin.The filters would say: this is how watercourses look in this kind of terrain.
Then you would switch to your normal Terrain Brush, the Terrain Protection slider would include the rivers in its influence because the rivers would be part of the terrain anyway. Or maybe you could have a Rivers Potection slider so that you didn't 'mess up' the river network?

monks

[ysvry]

looking forward to rivers , turning the problem around and starting with rivers seems a good extra option.

[LightDrop]

Very interesting points of all about river-creation in a Ter-generator. But the most important point in my eyes is anotherone - practically all render-software (except mojo with its river-system) offer just water-planes, which means all of the water is on a definated height! and of what use are the most sophisticated river-systems in the ter-generator, if you are not able to fill it accordingly with water in the render-application? If I'm wrong with this point, I'm more than happy to hear about!

c u

[cajomi]

You are absolultly right. And so, the river tool will generate a second terrain, which holds the water.