DEM-less Topo -- Mountain Coast

[Peter Toth]

Hello Guild,

Another solid effort to produce semi-realistic topography without relying on DEMs, for I feel much too guilty when I'm plagiarizing Mother Earth.

This project was cooked up on Photoshop and Wilbur using my highly intricate and experimental procedure, which I still have yet to finalize and write a tutorial about. (A document that may well exceed 20 pages, just to give you an idea--and STILL the topography fails to impress/satisfy me!)

Please criticize, and let me know if this is an improvement over my last few submissions.

Thanks,

Peter

Mountain Coast.png

[Cookiegod]

Well overall the map looks good. If you absolutely need one thing to criticise it'd be the lack of fault lines. The mountain ranges don't have a direction. That might have something to do with (I'm guessing now) you probably made the topography by rendering clouds and then generating a height map from it (with some steps in between). The map isn't bad in my opinion, but again, if you're dissatisfied, that might be why.

[Redrobes]

The map looks nice and is a reasonably realistic but you have a lot of lakes where there are inflows to them and no outflow. These sorts of lakes do exists but very rarely and are called Endorheic.

https://en.wikipedia.org/wiki/Endorheic_basin

As a general rule (which can be broken of course), but lakes have many inflows but usually only one outflow. If it has multiple outflows then they usually join back up pretty quickly. The reason is that in a stable and long time span, one of the outflows will dominate and erode where the others will silt up. In new geology like at fault lines or where glaciers have receded then things get more complicated and the process may have not evolved to the final state. Also, on river basins where there is a flat flood plain the river may channel around and leave pools but not like big mountainous hills with lakes between.

I think the issue is that you have created the landscape from some random process ignoring the effects of erosion and geomorphology. Its a good first approximation but these effects get modelled into the landscape over time and give that more realistic feel if you can manage to get them.

I believe that Wilbur has some basin fill options and maybe Waldronate can explain some of that better. But it can fix up that issue to some extent and leave you with longer rivers that run from the hills out to the sea.

It could also be just an effect of the processing where you have cut the map off at the right and bottom and the rivers are streaming out in those directions. Perhaps if the map extended a lot further before you cut it down its final size then the rivers would not outflow to the edges of the map. Its treating the right and bottom edges as sea at present which could be where the issue lies.

[Peter Toth]

Hello Cookiegod and Redrobes,

Thank you so much for the feedback. Yes, I believe I can now see what's going haywire with my topography: I used Wilbur's fractal noise generator to randomize the terrain AFTER I'd hand-drawn the mountains. I didn't realize this bit of processing erased the organic look of the mountains. So without further ado, here is the original.

Original.png

[waldronate]

A relatively quick way to get a DEM is to use a paint program to draw white blobs on a black background and then use the paint program's smudge tool to drag the black sea inland through the coast to form rivers. As the smudge progresses, it blends the black onto the white, running the "canyon" uphill. If your paint program has brush dynamics and can make the brush smaller as it progresses, the rivers will even get naturally less wide. Then plop that sucker into something like Wilbur and it's pretty easy to get something reasonably plausible after suitable processing (and usually an exponential operation).

You can also draw blobs with solid black rivers and use that as an input into Wilbur, but you'll need way more aggressive processing.

[waldronate]

... you have a lot of lakes where there are inflows to them and no outflow.

I suspect the color scheme and excessive river detail are tricking your eye here. There are also some rivers that run away from the coastline, which is a somewhat unexpected result of the how the map is cropped.

[waldronate]

My major complaint is that, again, you didn't specify a scale on your map. Some features are only plausible at a certain range of scales and some generators (Wilbur being a good example) only generate plausible results at a certain range of scales. If you generate a map with Wilbur at the wrong scale, it's going to look implausible.
Also, if you're going to start with hand-drawn anything, you want to let the procedural generators elaborate the high-frequency details (the small stuff). However, you need to carefully control the small stuff, especially with Wilbur's fluvial erosion things. If you've spent a lot of time drawing things by hand, only do a little bit of final work using the procedural generator.

Two major artifacts in the first image are apparent: rivers that run along 45 degree intervals and lots of rivers that run parallel before converging into a crow's foot configuration. Both of these artifacts are due to limitations of Wilbur's algorithms. The 45 degree things are because it runs on a square grid and finds flows only between corners. The crow's foot rivers are because the basin-fill step commonly used to establish overall flow patterns results in relatively flat areas with a single outlet. The flat areas let the precipitons meander around without pushing enough sediment to merge the rivers as would quickly happen in the real world.

A way to get rid of (or reduce) the 45-degree pattern is to use a multi-resolution process. Start with a low-resolution input to establish the major elements, process it, then scale it up, adding more white noise (not fractal noise) at each scale. The scale-and-add process generates a fractal noise on top of your low-resolution input image and the river-enforcing steps at each scale ensure that rivers do semi-plausible things. The hard part is to get enough white noise to force the rivers away from grid boundary alignments while not entirely overwhelming the rest of the landscape.

The crow's foot problem is much harder to get rid of. Rivers should generally merge in a hierarchy rather than all at once. Wilbur doesn't have any way to do that, so you'll probably need to put in something manually to break up the effect when it happens. A way to get nearly-parallel rivers to merge is to save a copy of the current map, blur it to force rivers to merge, make a river flow map (black and white), and save that flow map. You can tweak the flow map in an image editor if you want at this point to reduce some of the excess branches. Then load up your previous (unblurred) map, load the flow map as a selection, and force those rivers into the surface by subtracting a constant value. Deselect and run a little extra fluvial processing to smooth out the edges and it may help a bit.

I haven't found a solid way to get an unguided plausible river network in Wilbur from a random terrain due to the above two artifacts. If you've got a good hand-drawn flow network, Wilbur can work nicely for elaborating some details. If you're starting with an implausible network, Wilbur can happily elaborate those implausible details and add a few of its own. Wilbur is ancient and cranky and there are better (costlier) tools out there these days.