Wilbur making coastal cliffs

[MrBragg]

I've been experimenting with Wilbur a bit lately to process a starting rough heightmap and have been having a hard time getting Wilbur to preserve some of my coastal lowlands. Basically, things behave pretty well where I have large-scale regions of lowlands near the coast; there they start out low and stay low through 2 - 3x cycles of precipiton erosion. Where the coastal plain is more narrow though (e.g. where coastal mountains meet the sea, leaving only a few pixels of lowlands), the same number of precipiton cycles tends to fill the lowlands up and create cliffs instead of a smoother descent to sea level; this happens whether I have the land selected with a mask or not. I guess my question is whether there's a way to preserve these coastal lowlands while doing a few precipiton steps, or if the best strategy would just be to manually paint down the coasts at the end? So far the only thing I've found that works is to start off by doing a morphological erode, which tends to give a uniform "ring" of low terrain on the edge.

Any input is appreciated!

[waldronate]

What you're describing can be caused by the selection clipping what would be a smoothed area. If you feather the selection a bit (say, 0.5 to 1.0) before doing the precipiton erosion step, it will get an interesting bathtub ring. If you invert the selection and set the sea to a small negative value (-1 to -10), the landscape will get cut back from selection somewhat and you'll want a height clip operation from 0.01 to 1000000000 in order to force things back to the original selection.

It can also be caused by how the precipiton algorithm works if you've only got a little ledge between a high and a low place. The algorithm moves Delta amount of the difference between the high and low terrain in the downhill direction. For example, a sample at 10 and a sample at 1010 will move Delta*(1010-10) altitude onto the 10 pixel. With a default Delta of 0.25, that would be a change in altitudes from (10, 1010) to (260, 760). This process continues moving altitude downhill. Small altitudes will get overwhelmed pretty quickly if they are next to a large one. Transferred altitude that goes outside of the selection (or off the map edge) just disappears. The default Delta of 0.25 is the largest value that will do the right thing. Using a smaller Delta will probably give better results in this case, but it will take longer to get similar results.

[MrBragg]

That explanation make a lot of sense to explain what I was seeing. Feathering the selection gives much more control over the coastal altitude than I had before, so thanks for the fast--and detailed--suggestion / explanation!