Here's an overview, compiled from several my recent replies to interested developers.
***
Existing data from other architectures cannot be used easily (neglecting the legal side for now). Outerra comes with entirely different architecture and a different mapping of terrain data.
As you know Outerra engine uses procedural techniques to refine the rough geometry down to centimeters. It also uses procedural techniques to generate/mix terrain textures using climate data, local topology, and land class data. However this isn't exactly the same as in FS. Land class data will be derived from 500m data describing the dominant vegetation or surface, augmented possibly by data retrieved from analysis of Blue Marble data that would allow for finer transitions between climatic types.
All this is oriented towards reconstructing the Earth in its virgin state - as if without the touch of civilization. The effect of civilization will come as a separate layer, but it's meant to be predominantly based on vector data. The reason is that the "natural" base layer must be processed into a format suitable for further procedural refinement, so that the detail can be consistent across the whole detail range - from high altitudes down to the underfoot level, while at the same time it should follow a natural-looking fractal distribution that makes it so believable.
Modifications of the resulting natural base layer are then possible using vector overlays - basically you'll define a polygonal area, assigning it a land class, and the engine applies it as it creates terrain tiles. This allows for creating sharp or precisely defined transitions between land classes that will be properly divided by roads and rivers.
The engine is capable of rendering/generating consistent detail whether you are high up in the sky or standing on the ground. However, the direct consequence is that we aren't supporting the use of raster imagery (...) - it is not refineable, has lighting issues etc etc. Although it's possible to implement, due to the special mapping we use it would have to be reprocessed before it can be used, but this is true for all data used by the engine - the reprocessing also compresses the data into a LOD-friendly format suitable for streaming.
But in addition to zooming in looking terrible, the transition between vector+fractal defined regions and image mapped ones would have to be solved as well.
The use of vector data brings many advantages - for example much less bandwidth is used for upload on GPU, while GPU has a much higher internal bandwidth available than CPU->GPU bridge and is powerful enough to generate raster data used for actual rendering.
A disadvantage is that the vector data have to be obtained or retrieved first. Sources such as OSM can be used without licensing issues as the vector data are separate, and won't infect the remaining data sets with their license. The workflow for individual users could be such that if one wants to refine a previously undefined area (containing only the natural layer so far), he would be doing it in OSM, with a subsequent import bringing it to the Outerra world. For commercial addons the files with custom vector data definitions will have to be kept in separate files to avoid potential conflicting license terms.
Our ultimate goal is to make a universal simulator platform that combines flight, rail, road, sea and space simulation. The architecture definitely allows it, but it requires a new toolchain for making the scenery and new data.
***
We are creating FAQ section that will discuss these things in more detail, and I'll be doing a blog about it as well, that should visualize certain its aspects. I also expect we'll be getting more feedback on scenery creation then, once the architecture is better understood.
