Misaligned Roads in the city: Loei,TH

While mapping in the city Loei, TH we have observed that a good part of the city was not aligned to the OSM GPS Traces. And also identified that at a couple of locations the mapped data is not in sync with existing data - Part of the area is matching to Bing and another part is matching to Maxar/GPS Traces.

Note: The OSM GPS data is matching the latest imagery (Maxar) in this city, but the existing highway data were aligned in Bing in most of the regions.

Case1: Existing Data is in Offset with the latest imagery and OSM GPS Traces.
Location: 17.2928048, 101.765539

https://www.openstreetmap.org/way/672793480

Refer: https://www.openstreetmap.org/way/672793475

Case 2: Part of the area is matching to Bing & other to GPS Traces/Maxar.
The partial region has no sync in terms of offset with reference to per Maxar Imagery, which is on the right side of the water body (https://www.openstreetmap.org/way/461500428 ) and the left side area of the water body is completely misaligned as per Maxar Imagery
Offset Location: 17.3001042, 101.770402 , Matched Location: 17.3005708, 101.7782773
Reference WayID :https://www.openstreetmap.org/way/106994813

Few other reference areas where the issue is identified:

  • 17.3001255, 101.7689874
  • 17.3098093, 101.7647249
  • 17.299109, 101.7677489

We need your help and suggestion to fix the data at this location in Thailand and looking forward to your opinion on - whether to re-align the existing data to match with GPS Traces/Maxar Imagery? Or any other suggestion to sync the data across the city.

Ref: https://github.com/GRABOSM/Grab-Data/issues/109

This is a common problem. You should trust your judgement and move the map roads to agree with the GPS traces. It can be tedious as you will need to move a lot of intersections. Once you have an area clearly oriented, you can also move other nearby roads and features. This is an important improvement to the map, many thanks.

With space bar+left mouse you can move the entire map to agree with your GPS traces. At least with Potlatch.

Be aware that the map GPS orientation can change abruptly as you move into other areas.

A couple of years ago, I was in Loei, and afterwards did some edits. I detected the misalignments, too. But the situation was even worse: the offset differed severely between different areas of the town…

I would definitely say yes.

Context: Bing has been the default layer in ID, and it has been used primarily by new mappers due to a lack of alternative imagery and missing knowledge on how to change layers in editors. Initial mapping of cities was done primarily using Bing. I believe public GPS traces at the time was rare or inexistent.

Unfortunately, in Thailand, Bing seems to have most of the time outdated and misaligned imagery.
We have discussed this in the forum and documented alternative recommendations in our wiki.

Hello jinalfoflia,

as others already mentions: Please improve the data if you have better sources.

Please do this in a professional way. You can’t simply draw a bounding box to select all elements and move them. Some Elements might have already a proper alignment. You can use the bulk-move in case you select elements which the same mapper added in a similar time-frame. Here it is likely that all face the same offset.

You might also need to check the changeset comments of edits. In recent edits often the imagery source is recorded.

Please also consider that alignment problems in imagery can change and differ on various areas. GPS tracks can be off as well. It might be reflexion on buildings or low signal quality in urban canyon or forest.
Another problem could be if SBAS is used in Thailand, which can lead to false correction.
Example:

https://www.technologyblog.de/2012/12/warum-sbas-fluch-und-segen-zugleich-ist/

So GPS is only trustworthy if you have multiple tracks matching the imagery. And if you have alignment problems not by a linear offset, but due to problems with accuracy of the elevation model, then it gets worse again.

So please make sure that the data after correction is significantly better than before. The quality needed in practical use-cases is relative position. If you search for a doctor’s office, you need to know whether it is on the left hand side of a junction or on the right. It does not matter if there is an offset of 200m. Humans do typically orient on relative landmarks. So position of features relative to landmarks has to be right.

Routing engines can typically core quite well with such offsets, as it matches you on a position in the map data, which levels out the offset.

To be specific again: It does not help at all if you move the roads to match the exact GPS position if as a consequence all single nodes in the area do no longer align correctly relative to the other features.

Thank you everyone for the suggestions. I agree that this particular area has a different offset for different areas and the GPX traces aren’t present everywhere, we are finding better ways for us to align this city and will keep this thread updated with the outcome. We will need some more time to analyse and make the edits in this city.