Background 1: The organization I work at creates and maintains the road network of the entire country as line features, along with some points of interest. They do this by acquiring high resolution color imagery for the country and digitizing over it. When a new updated satellite imagery is acquired (from Airbus or Digital Globe), the digitized features look shifted.

Background 2: Being a frequent use of the Waze app, I sometimes use the waze editing tool to update streets, landmarks, parking lots etc. The Waze editing tool provides google maps to digitize over and update features on the app. When google periodically updates its satellite imagery, all the digitized streets and features are shifted. This shift doesn't affect the usability of the waze app, but when using the waze editing tool, can seem to indicate the digitization was done incorrectly.

The two part question I have is:

1) Are two satellite images of the same area always expected to have a shift because of the way satellite imagery is captured, processed and provided by imagery providers?

2) Is there any solution that the imagery customer can take or the imagery provider can take to ensure every imagery of a given area does not show a shift?


1 Answer 1


First - the overly broad answer. Yes - objects shift over time due to continental drift. However, that is not what you are seeing in your data (I hope, for your sake, as that would likely result in some massive earthquakes in your area).
What you are likely seeing is inaccurate georeferencing.

Highly accurate and automated georeferencing is not easy. The operators (DG & AB) try to be consistent and reliable, but with low accuracy elevation models, the orthorectification of data is always a bit unreliable and when getting data from two sources, the methodology may not even be the same.
As such, you should treat the georeferencing as somewhat dubious and adjust the georeferencing to match known features.
Additionally, if you have the data & technical capabilities, you should do your own orthorectification with a more precise elevation model than what the operators have access to. This does mean that you need to buy data at a different processing level from the operators, as it is not advisable to try to orthorectify data that has already been partially orthorectified. This data is often referred to as 'ortho-ready standard imagery', when purchasing it.
If you do your own orthorectification against a high quality elevation model, and well known control points, the shift between datasets will be minimal (usually within subpixel accuracy if everything is done correctly).

  • Thank you for you answer. To clarify my question for others, I wasn't referring to continental drift. Also, for generating the data, we don't use both operators (DG & AB) for the same project. For the project that creates data of the street network, we only use AB.
    – Thomas
    Oct 17, 2018 at 10:11
  • It's also interesting to hear 'Highly accurate and automated georeferencing is not easy'. So I take it that Google imagery shift is also due to their inability to get the georeferencing right?
    – Thomas
    Oct 17, 2018 at 10:13
  • Yeah - Google also tries to do it, but they have the same issues as everyone else, with regards to lack of good elevation information. They try, and they also pull in high accuracy data from national agencies, where applicable, which helps with the geolocation. Oct 17, 2018 at 10:27
  • I need to open my survey books, but continental drift (speed) is measured in cm/year. so you can imagine how slow it is. I've only seen it taken into account only in VEEEEEEERY precise measurements.
    – nickves
    Oct 17, 2018 at 11:31
  • Exactly, @nickves - I only pointed it out to avoid cheeky comments about it. Effectively, it doesn't have an impact, when working with VHR satellite imagery, which only really exists since 2000. Oct 17, 2018 at 14:38

Not the answer you're looking for? Browse other questions tagged or ask your own question.