Lateral movement of mapped data

I have been examining OSM for a potential African project and have been amazed at the amount of effort already put in by other editors.
However, in one area where I zoomed in to review the digitising to a scale where housing has appeared, the mapped data is displaced by about 10-12 metres.
All the mapped data is correct in relation to each other element, so it isn’t ‘bad’ cartography.
If I start to add in more detail, it is not going to match the old.
Any ideas how this has happened?
I can only guess that the ‘original’ satellite image used for digitising has been superceded and the new image has been georeferenced differently.
So what is correct?
The new image or the old data?
How can I check?
Or am I pushing the accuracy just a bit too far and that with ‘free’ satellite data, an error such as this is acceptable; but whatever the answer, I still have the problem getting my new information to fit the old.

Most/all the editors allow you to adjust the imagery offset to correct that type of problem. The trick is knowing how much and in what direction the imagery needs to be adjusted.

For that you need some on the ground references which is typically done by gathering GPS tracks in the area. Or, if you are not in the area but someone else has been, use the GPS tracks that they have uploaded to OSM. Assuming the tracks are along one or more roads or paths that can be seen in the imagery, line the imagery up to the tracks.

Satellite imagery offsets are not uniform, so if you have lined things up in one area don’t assume they are lined up a kilometer or two away. You will need tracks in all the areas you are mapping to be safe.

All that said, many times there are no GPS tracks to align to and the new imagery does not match the offset of the old and you don’t have anyway of knowing which, if either, is correct. I don’t know about your area but in my area I am finding the newer imagery is often much better aligned than the old so given the choice of adding buildings, roads, etc with an offset, I’d use the new imagery.

My preference is to avoid blindly shifting OSM data to match aerial photography, unless you have good reason to think this will be an improvement. The person who mapped it originally might, for example, have used carefully averaged GPS waypoints to establish a reference significantly more accurate than the image alignment. On a couple of occasions I’ve had well-meaning mappers move a whole area I’ve mapped away from its true position by several metres!

I’ve also seen the imagery at different scales have significantly different alignment, not always with the higher resolution version being the better one.

Though I do agree with n76 – newer imagery seems usually to be better. Perhaps contact the original mapper and ask?

Thank you for the useful information. I am more used to UK where it is easy to check.
From your advice I will add into the spec. that when we do ground verification we will use GPS to anchor what we have extracted from the imagery.
I’ll see what GPS has been uploaded already to see if that will tell me what has moved.
I didn’t realise I could move the imagery. Sounds risky - I hope there is a CTRL Z feature as well ;-))