[cossatot]

This is cool, thanks for posting. I especially like how there are several styles of coastlines.

I'm really interested in algorithmic generation of topography, probably by first creating fractal stream networks (either from the coast/ low elevation boundary or from watersheds/drainage divides). I'm a geologist so most game-oriented topography generators look super unrealistic to me, but making stream networks and working from there would be helpful with quantitative landscape analysis and simulations. Anyone heard of anything like this? If nothing else, then some way of filling in a polygon with branch-like fractals?

[hyperpallium]

Yes, this is one of the main approaches when I was looking at this. Sorry, don't have references at hand, but recall comng across a few when searching for "hydraulic" erosion and river formation.

IIRC, it took a little while to find the favoured search terms. I was using http://citeseerx.ist.psu.edu/ and http://scholar.google.com.au/

There's an interesting review paper by Musgrave (decades after he was lead author on the most cited paper on fractal-landscapes, erosion and rendering) where he notes what you do: fractal landscapes look uncannily realistic to most people but they really aren't very realistic. https://www.classes.cs.uchicago.edu/archive/2015/fall/23700-...

BTW Can I ask you something please? I may rave on a bit, sorry... I'm interested in stream formation by erosion, instead of your more feasible stream-first approach, but I'm concerned that I end up with canyon-filled mountains, and silted up seas. I need a countervailing process, for mountain-formation and sea-emptying, which is surely plate tectonics: squashing mountains up, and subducting seas. Do you have any thoughts/references on modelling this? From my research, the cause of plate movement is not quite well-understood, with tentative theories of roiling magma rising, and somehow also point sources (for island chains like Hawaii).

Wikipedia says the crust is thinner under oceans, because land is a better insulator and the sea cools it faster (but shouldn't that make it thicker?); and the weight of the sea pushes it down (but shouldn't land -- ie rocks -- be heavier than water, being denser, since they sink?)

Thanks for your thoughts!

[cossatot]

Thanks for the Musgrave reference, I'll check it out.

As far as your geological questions:

1) Silted up seas and canyon-filled mountains aren't actually that inaccurate. Mountains as we think of them are highly eroded, basically just the remnants of much larger volumes of rock that were uplifted. I'd ballpark a 'typical' mountain range (Rocky Mountain front range, Appalachians, Alps, NZ Southern Alps) at having 10-20% of the uplifted volume of rock remaining. Say the mean elevation of a mountain range is 1 km above the surrounding land, the peaks may be ~2 km above. The rocks that make up those peaks (granite, gneiss, schist) might easily have formed at 10 km depth, so ~9 km of rock was removed through erosion.

Note most of this erosion happens during rock uplift, not after! So it's not like the mountains were super high and then got whittled down. I mean that happens but it's less dramatic than you may think. So the 'stream first' approach is not wrong--the two processes are simultaneous.

2) Ocean basins are huge, and very deep, but there are very large accumulations of sediments there. The oceans are so deep not simply because the crust is thinner, but because it's more dense. Tectonic plates are basically rafts of rock floating on a more dense mantle, and ocean plates are ~10% more dense than continental plates, so they sit lower (Archimedes' principle). However there giant 'fans' of sediment where big rivers, especially those draining actively rising mountains, hit the sea. The Bengal fan, where the Ganges empties into the Indian ocean, is a good example. But the fans are tiny compared to the ocean basins.

In subuction zones, as you've mentioned, the sediment does get squeezed and thickened. Though the oceanic plate in a subduction zone may be 8 km below sea level, in some cases erosion of the plate, sedimentation in the 'trench' and tectonic thickening of the sediment may be enough that it fills the trench in completely and comes above sea level. Barbados is an example of this.

The sediments in subduction zones trenches (called 'accretionary prisms') do eventually get compacted and heated enough to become rock again and pasted onto the sides of the continent. Often after enough ocean plate is subducted, there will be some islands or microcontinents on the oceanic plate that are too thick and buoyant to get subducted and they will smash into the continent, and all the sediment in between becomes incorporated in between, and melted and metamorphosed enough to look a lot like the continents in between. This is how continents grow.

3) The causes of plate tectonics are very well understood--The mantle is convecting because it's much hotter near the core than at the surface. Oceanic plates are more or less the surface of those convection cells, and continents are light by-products that don't get subducted back around but get moved around and smashed into each other and pulled apart by the convection below.

There are also things called 'mantle plumes' that are less well understood but seem to be places at the core-mantle boundary that are much hotter and send up very localized batches of magma to the surface. These are small, relatively stationary, and independent of mantle convection, and so they basically perforate the plates as the plates slide by. Like a machine that spurts frosting onto cookies on a conveyor belt, except upside down. Or something. This is what Hawaii is.

4) I had a 4 but I'm cooking dinner and forgot what it was.

[jjp]

Thanks for taking the time to explain some geology/plate tectonics in an accessible way. HN is great for the random educational diversions away from plain old tech.

[roywiggins]

This might be up your alley:

http://mewo2.com/notes/terrain/

[cossatot]

This is a good reference, thanks.

[crisnoble]

I recall a most amazing screen saver that was installed on a mid 90s mac which would build mountains one layer at a time. Pretty far off topic but I have often tried to find that screensaver and this wonderful writeup triggered a memory. I would stare at that screensaver for hours as a child.