I’m not comfortable doing boundary fixes, but Evans, Georgia spills over into South Carolina in several places:
These little overlaps are pretty common. I usually use the TIGER place shape files for the state to decide which boundary is actually correct, unless someone has a better source.
1 Like
The city limits and CDP boundaries comes from TIGER 2008. The Census Bureau attempted to snap these boundaries to county lines, but they missed some:
There are some instances in the Census 2000 and Current County Subdivision files where subdivision boundaries appear to bleed into the adjacent county, as opposed to snapping to the boundary line between the two counties. The Census Bureau is aware of this issue and is in the process of correcting it. These corrections will be reflected in subsequent releases of the TIGER/Line Shapefiles over the next two years.
Our state and county lines originally came from a lower-resolution USGS dataset, which didn’t align with TIGER 2008 at all. Subsequently, many of these lines got conflated with rivers, because mappers often incorrectly assumed that boundaries follow river centerlines (thalwegs) and that no one cares where in the river the boundary runs. This appears to have been the case with the Savannah in changeset 58,689,112, though the previous USGS lines were so far off that they consistently ran aground on the South Carolina side.
At this point, I’d recommend dissociating the state and county lines from the waterway=river ways and realigning them according to the TIGERweb boundary layers. Afterwards, if any city limit appears to extend to the river’s centerline, you’ll probably want to conflate it with the state/county lines. For our purposes, it doesn’t really matter how far into the river a CDP boundary extends, as long as it stays in the right county, since CDPs are based on census blocks, which are basically convex hulls of households. There aren’t too many houseboats on the Savannah. But you can align it based on TIGERweb as well, just to make things tidier.
1 Like
This popped up for me because the OSM boundary for Evans overlaps the OSM boundary for South Carolina.
Yep, I expect that’s where you found it. I already dealt with all of the out of state boundaries that overlapped into Georgia by snapping them to the state line when I did a big cleanup over the last few weeks. There’s a ton of boundaries that are no longer accurate since 2008. I fixed the most-wrong ones but left a whole lot of “close enough”. Unfortunately the current state of boundaries in the map in a lot of places isn’t great.
I have been snapping updated civil and census boundaries to state and county lines where they coincide, but …
How much do we trust the current state and county boundaries? Would it be worth the effort to update them?
I took a look at matching the current TIGER/Line civil/census boundaries up with the current TIGER/Line state and county boundaries, and they don’t line up nicely. Part of the issue seems to be that the state/county boundaries in TIGER are mapped at a larger scale (lower resolution) than the civil/census boundaries.
I wouldn’t trust them at all.
When I’ve encountered this issue while mapping townships in Ohio, my workflow has been to cross-reference the USGS topographic map layer. These are old topographic maps, but boundaries tend to be quite accurate for the time period. If the TIGER boundary seems to vaguely follow the boundary in the topo maps, then follow the topo maps.
I was able to rely on the topo maps based in part on local knowledge about which township boundaries have changed since the maps were published. For county lines, you can also look up the county in OpenHistoricalMap to see if the boundary moved more recently than the topo map’s vintage. Relatively few county boundaries have changed since the 1970s or 1980s, so this is just due diligence.
1 Like
Well, which would you trust more, current TIGER boundaries for incorporated municipalities and CDPs, or current TIGER boundaries for counties/states?
Oh, those are different? I just assumed that the current generation of shapefiles all lined up. That’s messed up.
Yeah, they’re different. The Place (civil/census) boundaries mostly line up with the state boundaries. The county boundaries seem to wander more than the other two.
But all three disagree with historical USGS Topo, sometimes more, sometimes less.
We could look up NOAA’s coordinates for boundary monuments, but I imagine that would just give us another set of conflicting data.
Ladies and gentlemen, OSM stands at an edge of being able to offer the very best of our abilities where the TIGER / Census Bureau / newer / older (including USGS) data converge and are able to be entered as “good, better, best.”
Let’s strive to offer “best.” Yes, this can be substantial effort, so OK, offer that, or at least document what is known at any given point in time.
I suppose the other issue with the topo maps is that very often the boundary within a body of water will be labeled as “Indeterminate”. In other words, good luck! 
1 Like
I’m cool with “indeterminate.” Then no one can say the OSM data is wrong. 
2 Likes
It’s not ideal, but if it’s good enough for USGS, it might be good enough for OSM. But, hey, you got something better: “I’m all ears!”