This is a difficult issue that may need to be addressed on an international level, as this style of Buddhist temple complex may be found in many countries.
If we stick to the definition in wiki, it’s obvious that we should tag the worship building with amenity=place_of_worship and a temple grounds with landuse=religious, but we should be concerned about this issue because OSM was developed from a more Christian than Buddhist perspective. One thing, OSM mappers and users value an amenity=place_of_worship more than a landuse=religious. This is not a problem if we are thinking about a church or cathedral (as well as a mosque and even some Buddhist religious monuments), we are thinking of the building rather than the land, in this case a landuse=religious is useful for tagging the area around the church to define their boundary. However, if we strictly adhered to this standard, there would be a cluster of amenity=place_of_worship in the small area. Most of them would have the same name (Wihan, Ubosot, etc.) and would clog the map.
I looked through a taginfo for some information, and this is what I discovered. (Only way and relation)
religion=buddhist + landuse=religious = 1,457
religion=buddhist + amenity=place_of_worship = 27,064
religion=buddhist + building=* = 17,743
If we assume that all building=* are tagged together with amenity=place of worship (which is impossible), there would be at least 9,321 religion=buddhist + amenity=place_of_worship that are not buildings (There should be more, but I can’t find a way to combine 3 tags to see statistics.). As a result, the Buddhist temple grounds are currently tagged with amenity=place_of_worship at least 6.4 times greater than landuse=religious. Some of them may be tagged with both amenity=place_of_worship and landuse=religious as I’ve seen in Japan.
To keep things simple, perhaps we don’t need to identify a distinct place for amenity=place_of_worship and landuse=religious, because they can be the same! If we don’t tag the building as an amenity=place_of_worship and instead use the area, there shouldn’t be one area inside another, as that would be too complicated and difficult to find a boundary for each.
The suggestion to tag amenity=place_of_worship as a node solves one problem while creating another. It’s unhealthy in some ways since it doesn’t adhere to One feature, One OSM element principle. Furthermore, the temple area should be useful for some user’s purpose; separating them from their boundary may result in data loss. Also, mapping it as a node appers to be a kind of quickly mapped that will need to be cleaned up later.
If we wish to designate the complex’s center (which is not yet described in the wiki), we can create a proposal for a relation for Buddhist temple complex, which can provide a role for temple center (main building?). However, because the center is difficult to locate and somewhat subjective, and there is no physical evidence on the ground, simply not mapping them may be the best solution. OSM users can still use the temple mapped as an area for navigation. A GPS software should be able to take you to the temple’s gate and that’s it. If users want to go to a specific building, they should type the name of that building into the software to ensure it will take them there.
P.S. While the landuse=religious wiki page uses Asakusa Temple Complex in Japan as an example, ironically it is still tagged with amenity=place_of_worship in OSM (without landuse tag). 