I'd be *extremely* cautious of this value, as it is estimated from a perspective grid overlaid onto the 'overview' image, assuming that the whole 8.5km width is just visible in the mountains to the south, and following the approximate N-S alignment of the playable area, and the near edge being 'the end of the world'... So plenty of scope for getting completely wrong results.
I get an approximate *known* populated area of 5km x 4km within the 8.5x8.5km "world". A Sanity Check with the previously estimated 8.7 minutes for a north-south traversal at the gallop of this same extent gives a 34km/h speed which seems reasonable. It also fits my previous estimate of "around" 750m for the size of Novigrad, being around 3x the size of mediaeval Bath.
This is still quite a large area, and the game world is still the same size as the previous statement, but there is some wiggle room in exactly what things mean.
There are several areas where this 'guess' could break down.
The "far extents" could be a lower resolution than the 0.37cm!! of an 8.5km "middle bit" and aren't part of the world size calculation.
My perspective "guess" could be significantly off... I only have two visible 'edges' with no indication of scale and the knowledge that the near edge is "at least" on the image edge... lots of room for wrong interpretation here.
Photoshop "perspective" gives some odd results sometimes and some distortion was evident... enough to invalidate the whole exercise? Maybe.
Still these sorts of numbers sound plausible enough. (Water accounts for around 1/3 of that area-ish, but not all of it is "unplayable area" so it wouldn't be fair to discount the whole of this from the "world area", just as a note).