Buildings with unique IDs sometimes have identical locations

[tomay]

the “the_geom” field is showing duplicate values across multiple OSEIDs. In other words, I’m seeing different properties, within the same calendar year that have the same “the_geom” field.

In the data, it does seem like there are about 150 buildings that fit this query:

select * from (SELECT the_geom,count(distinct(id)) as id_count FROM public.seattle_buildings_2022_update_mr_20240425_v21 group by the_geom order by count(distinct(id)) DESC) as id_check where id_count > 1

Where the_geom has multiple building IDs associated with it

┆Issue is synchronized with this Asana task

[tomay]

Reviewing some of the buildings in that query, here's what I found:

• First record with a count of 16 all show from UW campus. No 2022 data here, and different street addresses, but all 16 unique building IDs have the same exact lat/lng

• Second example with a count of 8 records is all buildings with a name starting with University Village. It includes some 2022 data. Also somewhat different from the first example, multiple IDs, multiple building names (but all starting with "UNIVERSITY VILLAGE..."), but they all have the same street address.

• Third example (47.52593209 -122.3308402), no 2022 data, five building IDs and two building names (CLOVERDALE BUSINESS PARK (BLDG D), CLOVERDALE BUSINESS PARK ), same street address

• Fourth example (47.5829049 -122.3228994), no 2022 data, four building IDs, four building names (all similar to Airport Way Ctr - Bldg C ), same street address

• Fifth example (47.5309583 -122.3320685), no 2022 data, three building IDs, same name (all KENYON BUSINESS PARK), same street address

[tomay]

if the_geom is generated based upon lat/lon, then it would make sense that there would be duplicate the_geom values. Properties with matching addresses and matching lat/lon also makes sense, because the lat/lon are pulled from the city’s Geodatabase based on addresses.

I’m not super-concerned about the_geom ids overlapping. I think this is bound to happen because address assignment for benchmarking is a somewhat imprecise science.