6 replies to this topic

[MapMedia]

I had the understanding that US Census Tracts followed street, especially in urban areas, but apparently in Los Angeles County some do not.
Does anyone have an understanding why Tracts might not neatly follow streets?

Bonus Question: A good GIS method to accurately select tracts that do not follow streets.

Chris

[MapMedia]

I had the understanding that US Census Tracts followed street, especially in urban areas, but apparently in Los Angeles County some do not.
Does anyone have an understanding why Tracts might not neatly follow streets?

Bonus Question: A good GIS method to accurately select tracts that do not follow streets.

Chris

Thank you Hans for offering an excellent GIS approach - especially during Christmas break.

Here is a preview of a tract whose boundary does not follow streets, and am not sure what method it follows. I am interested in ideas / ways to account for these.
I have placed emails to ppl at the US Census Bureau but am keen to see if Cartofolk have experience with Tract boundaries.

Attached Files

•  LATRACTS.jpg   206.86K   47 downloads

[Hans van der Maarel]

Okay, I can understand if it follows individual property boundaries, but this one seems to cut right through buildings...

[mckirk]

Many census boundaries do follow street centerlines, and while those TIGER/Line streets meet National Map Accuracy Standards (I think for 1:100000 scale), the standards are not all that high: 90% of features within ~160ft. So, if you're working with street centerlines from another source (LA County?), the street lines are likely to be more accurate and probably won't be coincident with the census boundaries.

The census bureau is working to improve the accuracy of TIGER for Census 2010, though I don't know the status of the update. Check out the 2007 TIGER Shapefiles. I believe that the shapefiles include whatever accuracy improvements that Census has made, and might work better than the geometry that you used. If my experience is any indication though, they're still far from perfect.

As for the extra credit... you could find these errors by building a topology (though in ArcGIS this requires an ArcEditor or ArcInfo license) that includes the census polygon and street layers (polygons Boundary Must be Covered By streets).

Good luck.

[loximuthal]

The census bureau is working to improve the accuracy of TIGER for Census 2010, though I don't know the status of the update. Check out the 2007 TIGER Shapefiles. I believe that the shapefiles include whatever accuracy improvements that Census has made, and might work better than the geometry that you used. If my experience is any indication though, they're still far from perfect.

Others in the Bureau here are working hard to make these improvements for the upcoming 2010 decennial census. Happily the positional accuracy of our database is now vastly improved (we used to be mainly concerned with relational accuracy back in the day when practically nobody had a national geographic database). Sadly, I have to report, as an internal customer of that data making maps for Dress Rehearsal and related operations, a whole new crop of "curiosities" are now appearing.

Creating and maintaining an accurate, national database ain't that simple. I'm just glad it isn't my job

[MapMedia]

Thank you McKirk and Lothimuthal for the valued insight.

I spoke with TeleAtlas about their premium street GIS dataset and they assured me that they could also provide Census Tracts that follow their streets neatly They can't provide a sample or example, so am quite skeptical for now.

What I would like to do is remove ambiguity between tracts and streets as much as possible for an on-the-ground street level survey, which is using Tracts as their survey unit.

[Dennis McClendon]

It's not unusual to find tracts that follow streets and railroad tracks that existed in 1970 but not any more. This is a problem in my redeveloping Chicago neighborhood. However, that doesn't seem to be the problem with your LA County example.

The pre-1980 philosophy was that the census enumerators had to be able to follow the tract boundaries on the ground, so I'm puzzled by your example. You might want to look at a paper tract map of the area from the 1970 Census (in a big depository library in the area) to see if that sheds any light.