This is an important concept, this means that for any location in the world, the terrain is composed of a single texture corresponding to the discernable elements of that particular area. Try to imagine a large array of textures tiled over the extent of the landscape. Think that each of them is being specially designed to fit in that precise area, with all the needed details for it directly rendered onto the texture bits.

This means three things.

1) Detail can be high, as high as you can make the texels to look good, either by procedural texturing or by a skilful graphics artist.

2) The performance impact needed to render the full terrain after tessellation is minimized because it avoids the need for any multipass texturing schemes to occur, leaving more room for other dynamic special effects to leverage the hardware on a better way.

3) Effects of a static nature can be made permanently blended into the terrain itself, such as scorch marks, debris, growing resource fields and so forth. Again this avoids the need for any multipass decals and its corresponding overhead.

All this is consistent with a common reasoning by John Carmack, and others among the programming community, which states that for a given framebuffer size, a standard one pass scene only needs a bounded number of texels to uniquely texture every pixel displayed.

This means that no matter the levels of detail you throw into it the theorical end result is one texel per pixel, usually this is a bit more due to overdraw and perspective correction but what this allows is unlimited detail over unlimited viewable areas all done with a well known and bounded texel budget.

This is a valid and strong point in behalf of full unique texturing.