How do I erode my planet properly on FTPro/Wilbur?

[laevex_esre]

I have largely been using a2area's Israh tutorial to do all my erosion up until now and I was happy with it. But now I've loaded the real world data into FTPro, I've realised my map's inadequacies.

These two images are at the same scale, projection and latitude.

Earth:
Earth 0.1.jpg
Etarek:
Etarek 0.1.jpg

They look like completely different scales to me. Evidently my map still has a long way to go before it's as natural-looking as the real world. But that is my aim.

So how do I go about eroding my planet to this level using Fractal Terrains 3 and Wilbur? I really want to do this. I am fully determined to make my erosion look that good.

[waldronate]

The erosion model in Wilbur and FT looks most reasonable when an editing sample's resolution is on the order of a few tens of meters or smaller. It's plausible up to around a kilometer or so if you're careful. Much above that, though, and the results are unlikely to look particularly plausible. Getting that resolution on an FT map would require a minimum editing resolution of 40000 to get anything even close to reasonable. The 32-bit version of FT has a hard cap of 8190 samples, meaning that you're not going to get anything plausible-looking.

Note that an editing resolution of 40000 samples in FT would take on the order of 21GB of memory for working space, meaning that you're going to be swapping to disk continuously if you don't run out of swap file space first. With the swapping, I would expect to see days to weeks of 100% CPU runtime to run an incise flow operation on a surface of this size. That presumes that you could find a 64-bit version of FT to work with (at last check there isn't one publically available).

The same resolution arguments apply to Wilbur as well, with the exception that Wilbur does have a 64-bit version.

[laevex_esre]

Thanks Waldronate!

That makes me a feel a bit better because now I know it's pretty near impossible I don't feel like it's my fault.

Would you say the best way to get realistic erosion is to focus on smaller, regional maps then?
What sort of resolution/scale ratio is best? Something like a pixel per km?

[waldronate]

1km / pixel is about the minimum that you'll need to get plausible results. You will probably want to overlap adjacent maps by at least 10% to get plausible erosion joins.

Erosion is a global problem and not particularly amenable to breaking into tiles because the edge conditions really matter, esepecially incoming flows from other areas. For the programmer pedants out there, yes, the system could break the large surface down into arbotrarily many smaller tiles, but it would still be treating the tiles as a single surface for the purposes of the operation.