Ways that go no-where / incorrect tagging

Stephan / Johnny,

I wonder if you can help and do a batch edit where you search for any roads unconnected to the main highway network, where the source is digitalglobe, and delete them all.

All over Thailand, I’m seeing these bits of roads that are just in the middle of nowhere, with no thought given to the status being correct.

Examples abound of ways in the middle of forests, which are clearly agricultural tracks… they get the Unclassified road status, and without a surface tag, routing software will assume they are paved.
What a glorious cock up that will cause with the routing engines.

Minor roads often appear in the middle of factories, which should be Service, and should have a access tag of private or permissive.

The problem for me is using Potlatch and tracing from Bing, once a road is in there, it usually obscures the underlying surface. We all know with good aerial coverage, the best time to get the tagging correct is the first time … after that, mappers tend not to go over previous work.

Im changing to track or service or residential, where I can, but its a laborious job. Its easier to just delete a batch of roads when clearly the plotting has been automated with little concern for accuracy.

However, if you can do in bulk, that will be really helpful.

And I have sent a copy to Drishtie directly as I suspect this has a lot to do with that group !

Rgds, Russ.

Hi Russ,

selecting all the ways added by specific user, maybe even these from a changeset with a specific comment sounds easy.

For detecting unroutable “islands” there is already QA tooling in place. Have a look at OSMI. There are thousands of such unroutable places. Majority not coming from the Facebook edits, but I agree as professional paid mappers they should care more about the quality due to the large quantity of the edits.


I currently don’t know an easy way to detect unconnected roads from facebook. Might need some research or scripting.

DWG already offered FB to support in reverting their past edits. DrishT said they are going to manually review the past edits.
Whatever turns out to be the least work should be done.

Be careful with that tool. For example, I traced from Bing a village: https://tools.geofabrik.de/osmi/?view=routing&lon=105.56298&lat=15.68293&zoom=16 which is shown as a routing island. But the road leading to that village may not be suitable for “common” cars, and in some other discussion we decided to tag roads requiring 4wd as tracks.
Near the Satun ferry terminal, there are some villages which may require a ferry to reach them (or there’s some bridge which I did not see): https://tools.geofabrik.de/osmi/?view=routing&lon=100.04556&lat=6.56284&zoom=14
And in other places, clouds on the imagery may have obscured sections of roads below.
Consequently, I think that an automated edit could cause more issues than help.

The suggestion was to use an automated edit to remove the Facebook data.

Clearly there might be corner cases where the OSMI routing view does not detect a possible routing. In your second example the ferry connection is not mapped. So this is a correct report. Please map the ferry.

In your first example again I disagree with the tagging as a track. We had this lengthy discussion before. I still would tag this as unclassified and put surface=unpaved on it. The roads are typically tagged by function.

I note that some of your routing software does not provide the option to “avoid unpaved”, so it might you route on unpaved roads when not desired. Still tagging main roads leading to villages as tracks sounds wrong.
But lets keep this discussion out of this thread. We can go back to the older one if needed.

Here I still wait for a response from Drishtie on how they plan to clean up their previous bad edits.

The only automated edit sounding practical would be a complete revert of changesets. Either in full or only changesets we manually identify.

Hello Russ and Stephan,

Thanks for finding these. Russ would you help me with a few examples or are you able to share the usernames of the edits? As promised we will be completely responsible for our edits. I believe the edits from Facebook last year were all reverted but I would love to get to the bottom of these roads that go nowhere or not tagged correctly. Happy to have my team tackle these issues.

Our ID tooling also has validation checks for unconnected roads so editors cannot save work if there are any island roads with no connection. All roads that need to be connected in the data will be highlighted until fixed. We have similar checks for tagging as well.


OK, I have been fixing as I go, usually by changing the status, and adding roads as necessary to connect to the network.

The area that prompted this post was http://www.openstreetmap.org/?mlat=18.4789&mlon=99.6921#map=14/18.4789/99.6921 which was a nest of about 10 minor roads, clearly unconnected to the network, and almost all should have been tracks due to their location and usage.

I have fixed that area now, but clearly it provided an example of why Automated edits are so bad for the map if not done correctly. VLD002 and RVR006 were the culprits for at least one of the roads picked at random from the bunch, and clearly even our “poor quality” Bing image shows the way as unpaved.

If later digital globe imagery shows its been paved, I will apologize… but I ride off-road in Lampang a lot and I’m pretty sure their asphalting programme has not got that far ! This is a recent edit (Feb 2017), which I assume has passed validation ?? Im a little concerned it slipped thru DrishT, so then I wonder how much else has got thru ?
And remember I find these imports by chance because I know the area and something isn’t right… I’m not even looking for them, so I can’t comment on the magnitude of this issue.