osmlab/labuildings

Problematic assessor data on the beach

Opened this issue · 10 comments

I've found a lot of buildings on the beach designated as "house" or "industrial," but it turns out that their AIN is wrong.
screen shot 2016-04-24 at 6 42 19 pm
screen shot 2016-04-24 at 6 42 36 pm
You can search for an AIN on the county assessor's site. The example above shows that the AIN from the building file actually corresponds to a house deep inside the county.

I'm trying to figure out if there's a pattern or not. My first reaction is that the items on the beach aren't technically "parcels" at the county assessor. Maybe the building file defaulted to some way of pulling an AIN.

Shown location is here. That building isn't inside a parcel in the GIS dataset, so it does sound as though there's a problem with the coding here?

Mark Greninger, who works with L.A. County's GIS team, said that the building's center point was joined to the corresponding parcel shape and some nulls (like a building shape without a parcel) could get attached to other AINs. So, we may need to remake the building shapefile and kick out any lacounty:ain info that comes up for buildings not actually sitting on a parcel.

So, we may need to remake the building shapefile and kick out any lacounty:ain info that comes up for buildings not actually sitting on a parcel.

@jschleuss what's our next action here? Should we advise importers no to touch tasks on the beach until we remake the building shapefile? There are currently 3 untouched tasks.

@maning I'll jump in and look today.

Still working on this.

I've finally been able to figure this out. The problem isn't that widespread. (1,593 of 3,120,189 buildings or about 0.05% of total)

For these buildings I recommend we remove all tags that relate to the assessor. That's everything but building=yes, ele, height and lacounty:bld_id.

How do we do this effectively? We could fix those problem buildings and re-merge them with the file, re-chunk and all future buildings would be good. But we'd still have problems with those on the map. Any thoughts on how we'd fix those? Is there a way to search for a lacounty:bld_id in JOSM? We could create some strange tasking manager possibly (just a list of buildings)? Thoughts @maning

@jschleuss Can you send me a copy of the 1,593 buildings. I think we can manually review them easily.
Some thoughts:

  • Review the 1.5k buildings and edit the ones that were already by removing tags per your suggestion ☝️
  • For task blocks not yet imported we can assign it to ourselves. So we can fix it.

For a perspective of scale, 1.5K buildings is more or less two task blocks. So a manual review by experienced importers can be equally efficient.

Other ideas welcome of course.

@maning I like that plan. Also, I overshot the mark. While 1,593 buildings don't sit on a parcel boundary, only 929 mistakenly got an AIN. The others had NULL values. Here's the file: buildings-no-parcel.zip

I've kept it in the custom projection and left all relevant columns in the data.

Got this. Will update progress here.

That's everything but building=yes, ele, height and lacounty:bld_id.

@jschleuss
I am not removing start_date since this is from the original building data and not from assessor parcel csv.