What to do if GPS traces are misaligned. Do I adjust satellite image?

Hi OSM Community:

I’m a bit of a newbie, but I’ve started recording and uploading GPS traces using my ipad and the track-kit app. What I have found is the gps traces are generally accurate but overall it appears they are misaligned to both the map data and the Bing satellite imagery.

Here is an example:
I am using a large traffic circle as a reference to calibrate my gps trace. You can see that the gps is not only misaligned from the satellite image, but more importantly, the OSM map data is also off.

**What should change? (if anything) **

Option 1: Adjust the satellite background image. I am nervous about adjusting the satellite background to the map data since it may make all the data misaligned if it is incorrect. (is there a wiki article about the do’s and don’ts of adjusting the satellite background images before adding new map data?). In this case, adjusting the satellite does not reconcile the gps to the map data.

Option 2: Start changing the map data. I am willing to start ‘fine tuning’ the map data based on the gps readings, but given that an ipad gps reading may not be entirely accurate, I am just not sure if I am helping or if I would be making a mess of the OSM data.

I’ve checked the wiki and it talks about the accurace of gps readings, but I haven’t found specific instructions about interpreting my own gps readings and using that data responsibly for map editing.

Any help would be appreciated.


You can align Bing images, but I wouldn’t do that with just one trace. Turn on the GPS layer under the background settings in iD. When others have uploaded their traces for this area they should become visible. Then you can align the Bing images based on multiple traces.

I’ve heard that some Apple devices align the GPS tracing according to the road map in the device. see http://www.openstreetmap.org/user/Kachkaev/diary/20064 for some discussion on the topic.

I would also get in contact with the other mappers in the area to see what they did to align the images, or whether they just worked with the Bing images.

Hi escada:
Interesting article on the iphone. I get the sense that the gps in my ipad air is more accurate that this older iphone in the article, but who knows. It could be thrown off by Apple locking the readings to their maps … or it could be getting corrupted data for other reasons (surrounding terrain, roof of the car is blocking some satellites, etc).

Unfortunately, I’m getting gps traces for Kigali, Rwanda and there aren’t many other mappers adding gps here, so there is insufficient data to base a decision on.

I will keep doing what I’m doing: Adding gps traces and editing the maps as responsibly as I know how.
Thanks for the comment.

One other possibility - other GPS layers. There are a couple of options from Strava that might be useful - there’s a heatmap of points available from Strava cyclists and also one from Strava runners.

They’re available in JOSM’s imagery list, but you can also use them in iD too. JOSM displays them in its imagery list as:


At your location in the centre of Kigali there’s not much on the cycling layer but there is quite a bit on the running one. To see this layer in the iD editor go to “background” (the “stack of books” button at the right) select “custom”, “search”, and enter:


(i.e. lose JOSM’s “tms:” part and change “zoom” to “z”.)

FWIW you’ll see that the GPS tracks that are in OSM broadly match Strava’s on this occasion.

t is important to disable any feature in the device used to capture GPS data that uses information like an internal road map, as you otherwise end up copying the copyright map, not the GPS traces. If the traces look too good, you probably have this problem.

In most cases, real GPS data has errors that are larger than the errors in aerial imagery, however aerial imagery can have large errors in places, especially where there are height and therefore parallax changes. Even precision professional mapping, like UK Ordnance Survey maps can actually be out by several metres.

You therefore have to take each case on its merit and be prepared to use the de facto local datum for the map, even when you are sure it is wrong. If you are going to do major re-alignment, you need to contact the local mappers, who are probably rather thin on the ground in Rwanda.

The OSM map may be misaligned with your traces because:

  • the GPS constellation was in a different configuration at the time, resulting in a different error;
  • they were plotted from a government map which had errors;
  • they were plotted from a different set of aerial imagery;
  • they were plotted from a set of aerial imagery which has since been better rectified;
  • plate tectonics.

Actually, the discrepancy from GPS of both the OSM and aerial imagery alignments, in your example, seems to be well within the epected error range for GPS. As I said at the start, in most parts of the world, aerial imagery gives a much better location than a single GPS measurement. In that example, I would assume that the existing OSM was correct, re-align the aerial imagery onto that, use the GPS for the general topology and plot based on the aerial imagery. You need professional GPS surveying equipment, or GPS data integrated over a very long time, if you want really high accuracies from GPS.

The estimated error for GPS with the maximum number of satellites in view is officially about a 15m radiious circle, and for less than optimal configurations can be well over 50m