Currently I'm about halfway through GAZ (it takes a while as I launch a separate API query for every GAZ term.. would probably be faster to download all of Wikidata and do locally)
I have what I'm calling high confidence mappings for about 40% of the GAZ terms I see:
https://github.com/cmungall/environments2wikidata/tree/master/matches
These are unambiguous 1:1s (recall that many places in GAZ and Wikidata may share the same name)
I plan to do further processing of these to select ultra-high confidence (using parentage relationships)
This is likely lowballing the actual matches but it's a start, having a set of high confidence sameAs links will be tremendous for both resources, we can propagate geolocations, geonames xrefs, and ENVO classes for most of these!