TIGER data in OSM
TIGER is the USA Census Bureau’s street map dataset. David Hansen imported it into OSM at the end of 2007 and beginning of 2008. Took about five months. Lots of data there. Now, TIGER has a somewhat tattered reputation amongst the geo-cognoscenti and deservedly so. Take a look at this screen capture from a JOSM session northwest of Shippensburg, PA:
Notice how the error vector isn’t constant. It’s not like the data has been misprojected, which would result in it being off in one direction by a constant amount. Instead, the error vector wanders all over the place. It’s … as if the data was truncated at some point to a fixed number of bits or decimal points. So rather than all the points moving in the same direction, they all move to the nearest point representable by the truncation. I haven’t verified that that’s actually what happened, but that’s what it looks like.
Also, note that this is only true for some counties. Shippensburg is split between Cumberland and Franklin Counties. Cumberland is just fine, but Franklin is this horrible mess as pictured above.
June 26th, 2009 - Posted by Russ Nelson | 2 Comments
2 Responses to ' TIGER data in OSM '
Leave a reply
You must be logged in to post a comment.
on July 3rd, 2009 at 11:13 pm
This loss of precision theory might possibly explain TIGER errors within one small area, but there’s no consistency in the types of error seen across the whole US or even within one census region. Many errors are evidently the result of human error during surveying.
For example:
(map)
Generally the ways are not perfectly positioned down the middle of the roads seen in the imagery, but how do you explain the bizarre curly bit of road in the top right, or the deformed loop at the bottom? Just plain wrong.
on July 28th, 2009 at 12:48 am
I found out at the ESRI User Conference 09 that the TIGER data sometimes came from digitizing maps which were blown up too may times, so they were inherently inaccurate.