[Talk-us] Comparing Tiger 2017 dataset with OSM in a automatedway.

Kevin Kenny kevin.b.kenny+osm at gmail.com
Thu Oct 26 20:22:55 UTC 2017

On Thu, Oct 26, 2017 at 2:00 PM, OSM Volunteer stevea
<steveaOSM at softworkers.com> wrote:
> Thank you, Tod.  Yes, I MIGHT find a VERY SELECT SUBSET of these data SOMEWHAT useful, as minor amounts of them seem to be accurate and more-up-to-date enough to introduce into OSM.  But certainly not using any sort of automated method.  Essentially, every single datum would need to be human-reviewed, possibly corrected, likely conflated, and for a great many of them, on-the-ground verified.  I'd say "garbage" seems too strong, but "very noisy with a highly limited potential to add some minor value to our map, coupled with great effort to vet, improve and enter the data" seems about right.


An up-to-date TIGER is useful for
(1) identifying errors that the Census Bureau caught in the old TIGER,
and getting somewhat better data for the 40% or so of census tracts
that had stale data in the old TIGER because of some IT glitches at
the Census Bureau.
(2) (Maybe) Getting names for new features that are not in the map
yet, and building a 'things to map' list.

At least in the rural areas around me, TIGER, even TIGER 2017, has a
vaguely hallucinatory quality. It has much better data quality in
town, but where it is nearly correct all seems to be areas that are
pretty well mapped already.

All politics aside, where I have direct familiarity with TIGER, it's a
mess. It's a valuable data source when doing other mapping, a
semi-valuable data source when identifying things that may be missing
from the map, and would add considerable negative value if imported
without the process Steve describes.

All this is without any argument about whether imports are good or bad
for the community. It's unquestionably bad to import data that are
less reliable than what we already have.

More information about the Talk-us mailing list