First off, I want to say that I hate you guys for stealing my idea! Seriously though, I have long been interested in procedural terrain generation and often thought that the best way to fill in detail on a limited set of real-world terrain data would be to generate procedural refinement on top of that.
If I might digress a bit into theory, one thing that has always been seen as an obstacle in procedural terrain design is that earth's terrain is far from a perfect fractal. There are areas that are self-similar on multiple levels, like mountain ranges and certain coastal areas, but these are the exception rather than the rule.
Instead, there are large scale patterns that define continents, then smaller laws that define regions, and still smaller patterns that define even finer detail, all the way down to soil creep and rock weathering that gives you the exact shape of some four-foot cliff in someone's backyard. This is totally different from a fractal, which should follow the same rules no matter how deep you go.
This all means that you can't have one equation that defines "earth". This is where my idea and Outerra/anteworld's seem to have met. We can easily create the small-scale generators (well, more easily than other generators). It's easy enough to write the laws for sand dunes and grass cover and streambed erosion.
More difficult is simulating a few aeons of geological uplift and glaciation, and global climate shifting. However, these large-scale data are easy to extract from real-world data!
I was going in a slightly different direction- using study of earth-like fine detail generation and a crude approximation of large-scale generators to create realistic-seeming alien planets. I am glad to say there might not ultimately be any conflict in this at all- perhaps I might end up using the Outterra engine on my own project!
Now, about the Anteworld game...
EDIT: Looking over the forum, I realize a few things I mention here have been brought up before. Oh well.
I've spent a lot of time thinking about the story for the game. However, the more I think about it, the more I realize that perhaps the interesting part isn't the receiving, it's the telling. I think the greatest selling point of this engine is the fact that it's an entire virtual earth that's detailed down to the centimeter, not an epic quest or mystery.
As Anteworld is at least partially a means of selling the engine, and showing off its best qualities, why not focus on making the world as accurate to the real world as possible, in ways that will seem familiar to users?
The first thing people do when they load up google earth is find their house. Currently, such a thing is not really possible in Anteworld because roads aren't in. You can find the place that matches your local height map, but very few landmarks are present. If roads are there, though, suddenly you don't just have a height map- you have the power to race down your neighborhood street in a heavy duty combat truck.
Now, there are limitations to this approach- house positions are hard to extrapolate from data. You might find a way to interpret some kind of map, but the only data set I can think of that you could find that covers the entire planet would be something like population density and land use. You might even be able to find a 90m scale zoning map if you have access to a good source. Even if you combine all that with the road map, you won't be able to fool people. At best you could hope to put suburban low-density residential houses in a suburban low-density residential neighborhood with some consistency.
So if you want to trick people, you need an excuse. I say turn everything into rubble. Say there was a disaster of something, but it was only a couple decades ago. This is great because then you can plop down notable ruins, like a beat-up golden gate or the crumpled foundation of some Saudi vanity tower. A major city could have a downtown with a few ruined skyscrapers (possibly with internal meshes so we could climb around inside?) and some smaller buildings- enough so that if you're looking where you expect to see some famous skyline, you could see the ruined towers poking up.
As for what that disaster is, since we aren't looking at the entire population vanishing with all the buildings and everything, you have some more options. Maybe it was an asteroid storm or war, or just a plague followed by social breakdown. Maybe solar flares destroyed technology. It simply doesn't matter.
You'd ultimately be able to go to where your house is and immediately get a feel for the terrain. Then, when you go other places, even places you haven't been, you'll take that concreteness with you! The rest of the world will feel more real if you can associate it with a real-feeling place that you started out in or visited.
Going back to the story, as much as I understand the need to justify a lot of the construction and game tropes, but a high science-fiction setting might not be the way to go. If nothing else, it makes little sense to have a truck if your spaceship implies you could have a hovertruck. Why farm, if the ship presumably has all the supplies to keep a crew alive on four or five different alien worlds? If instead there is a lower sci-fi setting, with a few abstractions simply accepted for acceptance's sake, that may ultimately allow people to get more involved in the world. It's a fact of human nature that people would rather ignore buildings popping out of thin air than that a spaceship would carry thousands of people around in stasis for 2000 years and keep them on ice the whole time even though it's ostensibly taking them elsewhere.
Perhaps something more like fallout makes sense- you lived in a bunker while the disaster occurred on the surface, and now you rebuild. If a few truckloads of people need to appear out of nowhere or if it really shouldn't be possible to bootstrap an aerospace industry with nothing more than an iron mine and a crate of tools, well, we can get over that. It's better than wondering why the spaceship doesn't just land and get disassembled if we're really serious about colonizing earth.
Finally some miscellaneous questions:
The Moon? That is, is the moon going to be present as a visible object... and as a separate question, would we be able to, oh, you know, fly to it? Maybe it is there and I just haven't seen it.
On that topic, what about Mars? I don't suppose it would be possible to take martian height data and simply import it, or to invent a terraformed or partially-terraformed set of climate data for the planet either? This is rather more about the engine than the game.
Also, we are going to get freshwater too, right? I was kind of disappointed when I zoomed in on lake Tahoe and found a big grass plains... That said, river courses are going to be tough, because erosion is really hard to simulate...
Anyway, this post got a little long there, but as you can no doubt tell, I'm excited about this project.