I'm generally with NW_Kohaku on this, but with some minor differences. General thoughts:
* The largest DF map currently available is a "Region", and there are several different good reasons to believe that said Region is a part of a greater whole ("World", for simplicity).
* The largest currently available Region corresponds to a large island or small sub-continent in terms of practical extent.
* The lack of latitude variation, arguments about spherical vs. rectangular mapping, etc. are unfortunately in the scale range that it is not possible to unambiguously determine which effects are the result of design decisions, world shape, mapping conventions, simulation simplifications, and game mechanics. The simplest summary is that "the Region is a small enough patch of the World that the inaccuracies caused by representing as a square grid are low in comparison to other simulation issues".
* Regions are usually generated with a range of North-South climactic variation that seems unusually high for the practical size of the region if it occupied a similar size on Earth. This is somewhat in conflict with the previous point, as on our Earth these two assumptions don't go well together.
* The DF vertical structure is clearly quite different from that of Earth, in particular the presence of super-dense structure at the base of the geological column.
My default assumption used to be that the largest Region is a "trimmed octant", roughly 1/8 of a sphere with the pointy polar end trimmed off; but this doesn't work out as well as I'd like.
An interesting case could be made that the DF World is smaller and/or differently shaped than our Earth, but due to the presence of a hyper-dense basement the net surface effect remains similar (cf Ringworld and Scrith, for one vaguely similar example).